A collaboration between the Morgan lab at UCSF and the Gygi lab at Harvard has resulted in a paper by Holt et al. in Science, which reports the identification of several hundred substrates of the central cell-cycle kinase Cdc28p (also known as Cdk1) in the budding yeast Saccharomyces cerevisiae:

Global analysis of Cdk1 substrate phosphorylation sites provides insights into evolution.

To explore the mechanisms and evolution of cell-cycle control, we analyzed the position and conservation of large numbers of phosphorylation sites for the cyclin-dependent kinase Cdk1 in the budding yeast Saccharomyces cerevisiae. We combined specific chemical inhibition of Cdk1 with quantitative mass spectrometry to identify the positions of 547 phosphorylation sites on 308 Cdk1 substrates in vivo. Comparisons of these substrates with orthologs throughout the ascomycete lineage revealed that the position of most phosphorylation sites is not conserved in evolution; instead, clusters of sites shift position in rapidly evolving disordered regions. We propose that the regulation of protein function by phosphorylation often depends on simple nonspecific mechanisms that disrupt or enhance protein-protein interactions. The gain or loss of phosphorylation sites in rapidly evolving regions could facilitate the evolution of kinase-signaling circuits.

The paper makes several interested in analyses and observations. However, I found the comparison to the previous study of Cdc28p substrates by Ubersax et al. from the Morgan lab to be less detailed than I had hoped for:

Phosphorylation of Cdk1 consensus sites was observed on 67% (122 of 181) of proteins previously identified as Cdk1 substrates in vitro (4). Sixty-six percent (80 of 122) of these proteins contained sites at which phosphorylation decreased (log₂ H/L < –1) after inhibition of Cdk1 (only 45 of 122 are expected if there is no correlation between the experiments in vitro and in vivo; χ² test, P < 10^-10).

In other words, 44% (80 of 181) of Cdc28p substrates identified in the old study were confirmed by the new study, and only 26% (80 of 308) of the Cdc28p substrates identified in the new study are supported by the old study. There are many possible explanations for this discrepancy

Depth of the mass spectrometry

It is notoriously difficult to identify peptides from low-abundance proteins in mass spectrometry. In the new mass spectrometry study, the authors were able to map 8710 precise phosphorylation sites on 1957 proteins. However, budding yeast is estimated to express in the order of 4500 distinct proteins during exponential growth (Gavin et al., 2006). Assuming that the majority of these proteins contain sites that are phosphorylated during at least part of the mitotic cell cycle, it is likely that a considerable number of low-abundance Cdc28p substrates identified in the old study have been missed in the new study.

Biases in phosphopeptide enrichment

When doing phosphoproteomics, it is necessary to first enrich for phosphopeptides to improve the coverage. To this end, Holt et al. used immobilized metal affinity chromatography (IMAC). In 2007, the Aebersold group at ETH published a paper showing that different purification methods lead to isolation of different, partially overlapping segments of the phosphoproteome. Specifically, they showed that IMAC enrichment biases the data towards isolation of multiply phosphorylated peptides. Given that only a single purification method was used, it is likely that in vivo Cdc28p substrates may have been missed in the new study, in particular if the peptides contain only a single phosphorylation site.

In vitro vs. in vivo conditions

The old study by Ubersax et al. was done performed on cell lysate, which is an in vitro strategy (although all other proteins expressed during the cell cycle are present). It is thus likely that some of the proteins that are phosphorylated by Cdc28p under these conditions are nonetheless not in vivo Cdc28p substrates.

Can we do better?

As always, it is easy to point out potential flaws in other people’s data sets; however, it is much more constructive to do something about the problems. The challenge is thus to construct a larger and more reliable set of Cdc28p substrates by combining the data from the two studies.

To check the feasibility of assigning confidence scores to different putative Cdc28p substrates, I tested if the fold change observed in the new study correlates with the chance that the substrate was also identified in the old study. To this end, I divided the 308 Cdc28p substrates from the new studies into two groups and constructed histograms of the fold changes for each group:

The fold changes are clearly skewed towards larger negative values for the Cdc28p substrates also identified by the old study relative to the proteins that were not previously identified as Cdc28p substrates. This difference is statistically significant at P < 1% according to the Kolmogorov-Smirnov test. This suggests that the observed fold changes in the new mass spectrometry study correlates with the likelihood that the proteins are true Cdc28p substrates.

The old study gave rise to so-called P-score for the individual proteins (not to be confused with P-values). I decided to test if these too can be used as quality scores, I constructed an equivalent histogram in which the Cdc28p substrates found in the old study were divided into two groups based on whether or not they were also found in the new study:

In this case, no obvious trend is seen and a Kolmogorov-Smirnov test indeed reveals no statistically significant difference between the two distributions. Surprisingly, the P-scores do thus not appear to be useful quality scores for the putative Cdc28p substrates.

Given the two sets of putative Cdc28 substrates, only one of which can be ranked by reliability, how can we create a better combined set? If one aims for the high accuracy at the price of low coverage, one could obviously choose to trust only the substrates identified by both screens. However, given the caveats regarding depth of mass spectrometry and biases arising from the enrichment procedure, I would be hesitant to use this approach. Alternatively, one could aim for maximal coverage at the price of accuracy by trusting all sites identified by either study. However, seeing the large fraction of novel substrates identified by Holt et al. with a log2-ratio only slightly below -1, I would personally tend to apply a more stringent threshold to the data from the new study by Holt et al., for example requiring log2-ratio below -2, before merging the sets of substrates from the two studies.

Cite this post

I guess it is no secret to anyone that Facebook as agreed to acquire FriendFeed. Several people seem puzzled why I left FriendFeed only 3 hours after learning this news. I can understand that this may look like a knee-jerk reaction, but there is logic behind the madness.

The truths is that my existing setup of Web 2.0 services was not working nearly as well as I would like. The sheer amount of content being shared on FriendFeed meant that it was easy to overlook a blog post from one of my favorite bloggers, for which reason I still subscribed to their blogs as RSS feeds. This caused me to waste time because the same posts appear in two place, and I could not filter out the blogs on FriendFeed because most comments would be posted there and not on the blogs. Receiving everyone’s tweets on FriendFeed tended to create a background noise that would drown all other conversation; however, I could also not filter out the Twitter streams on FriendFeed and follow people directly on Twitter instead because many cross-post all their FriendFeed “likes” and/or comments to Twitter!

Given the new situation, it was clear to me that the time had come to fix my broken social network setup and redo the plumbing in such a way that FriendFeed would no longer be responsible for gathering most of the content. Looking at FriendFeed, I discovered that most of the content of interest originated from just three sources: RSS feeds of blogs, Google Reader shared items, and Twitter. By following people directly on Google Reader and Twitter, both of which I was already using on a daily basis, I was thus able to relegate FriendFeed to a much less important role. I still feed my content from other sources into FriendFeed and I occasionally check for comments on my posts; however, it is no longer where I read content posted by others. Coincidentally, the new role of FriendFeed is almost identical to the role that Facebook has played all along.

To make a long story short, I’m not leaving the friendly community at FriendFeed in anger. I still read the content produced and shared by the same people as before. I have just fixed the plumbing.

Several large kinase inhibitor screens have been published in recent years. Two of the largest come from Stefan Knapp’s lab and Ambit, respectively. The former group used a temperature shift assay to measure the change in thermal stability of 60 human serine/threonine kinases that is caused by the binding of each of 156 kinase inhibitors (Fedorov et al., 2007). The latter group used a competition a competition binding assay to measure the dissociation constants (Kd) for 38 kinase inhibitors and 290 distinct kinases (Karaman et al., 2008).

The two screens are not directly comparable because one measures temperature shifts whereas the other measures dissociation constants. To see if it possible to convert temperature shift values to Kd values, I asked Damian Szklarczyk (who is a Ph.D. student in my group) to map all data from both screens onto a common set of chemical and protein identifiers, extract all inhibitor-kinase pairs that were measured in both assays, and make a scatter plot of -log(Kd) as function of temperature shift. The result was a set of 704 pairs of temperature shift and Kd values. In the plot below, inhibitor-kinase pairs for which binding was not observed in the competition binding assay were defined to have a Kd of 10 microM, and negative values from the temperature shift assay were treated as zero temperature shift.

The plot shows that the two assays are in very good agreement, which is surprising considering that the assays are fundamentally very different and were run using different expression constructs for several of the kinases. The linear Pearson correlation coefficient is 0.92 when excluding the one obvious outlier shown in red (BIRB796 vs. MAPK11; this appears to be a false negative in the competition binding assay).

The linear fit gives an intercept with the y-axis of 4.9223, which implies that a temperature shift of zero (i.e. no binding according to the temperature shift assay) does not translate precisely into a Kd of 10 microM (i.e. no binding according to the competition binding assay). We thus did a second linear regression in which we forced the intercept with the y-axis to 5 (red regression line in the plot). We thereby at the calibration function -log(Kd) = 5+0.244*Ts, which allows us to to convert temperature shifts to Kd values. We have thereby managed to put the measurements from the two kinase inhibitor screens onto a common basis that facilitates direct comparison and integration.

Full disclosure: I have an on-going collaboration with Stefan Knapp’s lab related to screening of kinase inhibitor.

Cite this post

About half a year ago, I began experimenting with Second Life as a tool for virtual conferences (I should add that my experiences have since improved). However, I believe that imitating real life in a virtual world is not necessarily the best way to use the technology – it may be better to use virtual reality for doing the things that are difficult to do in the real world. A good example of this is Hiro’s Molecule Rezzer, which is one of the best known scientific tools in Second Life. It, and its much improved successor Orac, allows people to easily construct molecular models of small molecules in Second Life.

After speaking with several other researchers in Second Life, who like I are interested in evolution, I set out to build a similar tool for visualization of phylogenetic trees. The result is SLIDR (Second Life Interactive Dendrogram Rezzer), which based on a tree in Newick format constructs a dendrogram object. The first version of SLIDR can handle trees both with and without branch lengths; however, I have not yet implemented support for labels on internal nodes or for bootstrap values.

The picture below shows an example of a dendrogram that was automatically generated by SLIDR based on a Newick tree:

There is a bit more to SLIDR than this, though. After the dendrogram has been built, it can be loaded with a photo and/or a sound for each of the leaf nodes. When click on a node, the corresponding sound will be played and the photo will be shown on the associated screen (the white box in front of which I stand):

I plan to work with collaborators in Second Life to construct dendrograms for evolution of bats (including their echolocation sounds and photos of the animals) and for the fully sequenced Drosophila genomes. Please do hesitate to contact me if you would like to use SLIDR on another project. I intend to make SLIDR available as open source software once I have implemented support for the full Newick format.

Cite this post

After months of hard work from the entire STRING team – thanks everyone -  I am pleased to be able to say that STRING v8.1 has now been put into production. Here is a screen shot of the start page:

This is a minor release of STRING, which means that the imported databases of microarray expression data, protein interactions, genetic interactions, and pathways as well as text-mining evidence have all been updated. We have also fixed a bug that affected the minority of bacteria that have multiple chromosomes.

Another notable feature of STRING v8.1 is the new interactive network viewer that is implemented in Adobe Flash:

For further details please see the post on the official STRING/STITCH blog.

Cite this post

It has been much too long since I have last written a blog post. Part of the reason has been that I have been busy moving back to Denmark, starting up a research group, and co-founding a company. More on that in other blog posts. The main reason, however, has been a lack of papers that inspired me to do the simple follow-up analyses that I usually blog about.

This has thankfully changed now. Pedro Beltrao and coworkers recently published an interesting paper in PLoS Biology on the evolution of regulation through protein phosphorylation. The paper presents several interesting analyses and comparisoins of phosphoproteomics data from three yeast species; the abstract summarizes the findings better than I can do:

Evolution of Phosphoregulation: Comparison of Phosphorylation Patterns across Yeast Species
The extent by which different cellular components generate phenotypic diversity is an ongoing debate in evolutionary biology that is yet to be addressed by quantitative comparative studies. We conducted an in vivo mass-spectrometry study of the phosphoproteomes of three yeast species (Saccharomyces cerevisiae, Candida albicans, and Schizosaccharomyces pombe) in order to quantify the evolutionary rate of change of phosphorylation. We estimate that kinase–substrate interactions change, at most, two orders of magnitude more slowly than transcription factor (TF)–promoter interactions. Our computational analysis linking kinases to putative substrates recapitulates known phosphoregulation events and provides putative evolutionary histories for the kinase regulation of protein complexes across 11 yeast species. To validate these trends, we used the E-MAP approach to analyze over 2,000 quantitative genetic interactions in S. cerevisiae and Sc. pombe, which demonstrated that protein kinases, and to a greater extent TFs, show lower than average conservation of genetic interactions. We propose therefore that protein kinases are an important source of phenotypic diversity.

Figure 1a in the paper shows the intriguing observation that, despite rapid evolution of individual phosphorylation sites, the relative number of phosphorylation sites within proteins from different functional classes (Gene Ontology categories) remains remarkably constant between species:

However, it occurred to me that this could potentially be a consequence of longer proteins having more phosphorylation sites, and protein length being conserved through evolution. I thus counted the number of unique phosphorylation sites identified in each protein (thanks to Pedro Beltrao for providing the data) and correlated it with the length of the proteins. In the two plots below, I have pooled the proteins so that each dot corresponds to 100 proteins. The upper and lower panels show the results for S. cerevisiae and S. pombe, respectively:

As should be evident from the plots, the average number of phosphorylation sites in a protein correlates strongly with its length, which is by no means surprisings. It is unclear to me why the intercept with the y-axis appears to differ from zero in both plots; suggestions are welcome.

The next question was whether the Gene Ontology terms that correspond to proteins with many phosphorylation sites are indeed assigned to proteins that are longer than average. I thus examined the terms “Cell budding”, “Morphogenesis”, and “Signal transduction”.

The average S. cerevisiae protein is 450 aa long. Proteins annotated with “Cell budding”, “Morphogenesis”, and “Signal transduction” are on average 1.6 (739 aa), 2.1 (945 aa), and 1.5 (679 aa) times longer, respectively. By comparison, the corresponding ratios observed for phosphorylation sites are approximately 2.3, 2.6, and 2.4. It would thus appear that differences in protein length between functional classes of proteins account for much, but not all, of the signal that was observed by Beltrao et al. when comparing the number phosphorylation sites.

Edit: Make sure to read Pedro Beltrao’s follow-up blog post, which nicely confirms that whereas protein length does play a role, it is not the full story.

Cite this post

Yes, it is that time of the year again – we are now almost three weeks into 2009, most papers published in 2008 have hopefully made it into Medline, and it is time to reveal the words of 2008. In other words, I have updated the BuzzCloud resource and here is the result for 2008 (click on the image to go to the web resource):

I am thrilled to see the outcome. Without any cheating or tweaking, several buzzwords related to proteomics make it on the list with “phosphoproteomics” and “quantitative phosphoproteomics” being the two most prominent of them. Nice for me to see considering that my new research group at the Novo Nordisk Foundation Center for Protein Research will focus heavily on improving and applying the NetworKIN and NetPhorest resources for analysis of phosphoproteomics data.

This blog has been very quiet for a long time. There are several reasons for this, most of which are positive: I have not had many boring or negative results to write blog posts about, I have been busy writing manuscripts about the positive results instead, and I have moved to Copenhagen where I am busy starting my own research group at the Novo Nordisk Foundation Center for Protein Research. There is also one more reason for the absence of blog posts from me: I have spent a lot of time experimenting with Second Life, and that is the topic of this blog post.

I first got interested in Second Life when I heard that Nature Publishing Group was setting up a virtual conference center called Elucian Islands. In the beginning I felt very alone on Elucian Islands. There was a good reason for that – I was alone most of the time. My view on Second Life was thus that it was pretty (see images below) but rather useless.

I obviously took a look at the SciFoo presentations (seen in the background of the image above) and the other scientific displays at Elucian Islands and elsewhere in Second Life. However, these mostly reinforced my negative view of Second Life being fairly useless, since almost everything I saw was already being served better by dedicated resources. For example, slide shows are much more conveniently viewed and shared in SlideShare than in Second Life, and 3D protein structures can be examined and analyzed better in programs such as PyMOL.

Over at FriendFeed, Jean-Claude Bradley fought a brave fight trying to convince me that Second Life is in fact useful for science. His key point was that Second Life is all about interacting with people, so I should try to go to some scientific events in Second Life. Sadly, there are still not many such events, and although they have changed my view on Second Life, they have also shown that there are many problems that remain to be solved.

The first virtual seminar I went to was “Cancer, Cell Cycle, and Check Points” organized by Digi S Lab. This was a perfect match since I work on cell-cycle regulation myself. The seminar consisted of two excellent presentations given by Letizia Cito from Sbarro Health Research Organization and Fayamdria Foley from the American Cancer Society.

Whereas the presentations were great, the seminar also illustrated several of the problems that need to be overcome before virtual conferences in Second Life are ready for prime time. When the first talk started, I could not see any of the slides. Restarting my Second Life client did not solve the problem, nor did a reboot of my computer. After giving up solving the problem, the entire region in which the seminar took place suddenly crashed causing speakers and participants to all be logged out. When it came back online after some minutes and everyone had found their way back, I could suddenly see the slides. Even then, however, they took so long to appear on my screen that the presenter had typically explained half of what was on a slide by the time I could see the slide. I see this as a major problem that must be solved before Second Life conferences can work properly – it must be possible to change slides without a noticeable delay.

The second event I went to was the “ESRC Complexity Research Seminar in Second Life” that took place at Elucian Islands. This seminar was very different from the one described above in that it was not a purely virtual seminar; instead it was a video feed from a real-world seminar that was being transmitted into Second Life. Think of it as a virtual overflow room – the image below shows the people who had gathered shortly before the event started.

Sadly, this event was marred by technical problems. The sound stream was of such poor quality that the Second Life participants could barely understand a word of what the speakers were saying, and the video stream was of too low quality to be able to read their slides. I do not want to dwell on this but just note that good quality microphones and cameras are a prerequisite for streaming events into Second Life.

The third event I went to was the “Virtual Conference on Climate Change and CO2 Storage“, which again took place at Elucian Islands. This was again a mixed event taking place both in the real world and in Second Life. The presentations were excellent and important lessons had been learned from the previous events. The microphones worked perfectly this time, and the video feed had been abandoned in favor of showing a copy of the actual slides in Second Life, which greatly improved the readability.

In addition to these events, Elucian Islands now also runs regular events such as the weekly Nature Podcast event where a fairly large group of people gather to listen to the latest podcast shortly after it has been released (image from Joanna Scott’s blog).

Regular events are crucial in SL because they bring people together in the same place at the same time. The need for people to be online at the same time is in my view one of the major drawbacks of Second Life compared to other tools that researchers can use for social networking. In my view Second Life should thus not be seen as competing with tools like FriendFeed or Twitter, which you can read when you feel like it, but rather as virtual reality alternative to video conferences. I think that Nature Publishing Group is on the right track with this, and I hope that the few remaining technical hurdles will be overcome in the near future.

Full disclosure: I have been working with the staff from Nature Publishing Group trying to solve technical challenges on Elucian Islands.

The latest issue of Science features a paper by Yu et al. in which they report the results of a comprehensive yeast two-hybrid (Y2H) screen for interactions between budding yeast proteins. Just a few months earlier, Science published a paper by Tarassov et al. that describes a similar screen performed using a novel protein fragment complementation assay (PCA). Peer Bork and I wrote a Perspectives piece on these two papers, showing that the different assays for detecting protein interactions are complementary in the sense that they capture interactions for different subsets of the proteome. For example, PCA detects many interactions for membrane proteins whereas Y2H detects many interactions for nuclear proteins.

As part of writing the Perspectives piece, I performed numerous analyses that were not included in the final publication, because they were either too technical for a broad audience, not interesting enough to spend valuable space on, or would involve additional figures. Thankfully, my blog imposes no limitations on the number of words or figures (nor is it required that the content is interesting, although that is desirable).

The comparison included, in addition to the two interactomes introduced above, a third interactome that consists of all the high-confidence interactions identified by Gavin et al. and Krogan et al. using the tandem affinity purification (TAP) method. Also included in the comparison (but not in the Perspectives piece) was the literature-curated (LC) set of interactions published by Reguly et al. in 2006.

The Venn diagram below shows the overlap of the four interactomes in terms of proteins, that is a protein is considered to belong to an interactome if the method in question suggested at least one interaction partner:

The numbers outside the ellipses specify the total number of proteins for which a given method identified interactions. Notably, the PCA, Y2H, and TAP interactomes cover only approximately one sixth, one third, and half of the yeast proteome, respectively, despite all three assays having been tested on all yeast ORFs. This suggests that only a fraction of proteins can be targeted with a given assay.

A second way to compare the four interactomes is to count their overlaps in terms of pairs of interacting proteins. To provide additional detail, I distinguished between interactions that are not found in a given interactome because one or both proteins are not covered by the interactome in question (dashed lines in the diagrams), and interactions that were not found despite both proteins being covered (full lines in the diagrams). The Venn diagrams below show all twelve pairwise comparisions of the four interactomes:

As expected, the largest overlap is observed when comparing the two largest interactomes (LC and TAP), whereas the smallest overlap is observed when comparing the smallest interactomes (PCA and Y2H). Even if taking into account the differences in terms of protein coverage, however, the the overlaps between the interactomes leave a lot to be desired.

There are several reasons for the poor overlap at the level of pairwise interactions. One is that false positive interactions are unlikely to be reproducible by a different assay. A second is that the assays measure fundamentally different types of interactions: PCA and Y2H measure direct binary interactions between proteins, whereas TAP measures co-complex interactions, that is whether two proteins are part of the same complex or not. This is illustrated in the figure below, which shows the binary and co-complex networks for three different scenarios:

The two types of assays have different strengths and weaknesses. Binary interaction assays can in principle distinguish between the two first complexes, which only differ in that the subunits B and C are in direct contact in first complex but not in the second. However, binary assays are not able to distinguish between the second and the third scenario, that is whether A, B, and C form a single complex (ABC) or two complexes (AB and AC). Conversely, data from co-complex assays are able to answer the latter question but are unable to distinguish between the two first scenarios. The different assays thus complement each other, not only because they are able to interrogate different subsets of the proteome, but also because they provide us with complementary information about the composition and topology of protein complexes.

Cite this post

Two days ago, Matthias Mann’s group published a paper in Nature in which they compare the level of individual proteins in haploid relative to diploid budding yeast cells:

Comprehensive mass-spectrometry-based proteome quantification of haploid versus diploid yeast

Mass spectrometry is a powerful technology for the analysis of large numbers of endogenous proteins. However, the analytical challenges associated with comprehensive identification and relative quantification of cellular proteomes have so far appeared to be insurmountable. Here, using advances in computational proteomics, instrument performance and sample preparation strategies, we compare protein levels of essentially all endogenous proteins in haploid yeast cells to their diploid counterparts. Our analysis spans more than four orders of magnitude in protein abundance with no discrimination against membrane or low level regulatory proteins. Stable-isotope labelling by amino acids in cell culture (SILAC) quantification was very accurate across the proteome, as demonstrated by one-to-one ratios of most yeast proteins. Key members of the pheromone pathway were specific to haploid yeast but others were unaltered, suggesting an efficient control mechanism of the mating response. Several retrotransposon-associated proteins were specific to haploid yeast. Gene ontology analysis pinpointed a significant change for cell wall components in agreement with geometrical considerations: diploid cells have twice the volume but not twice the surface area of haploid cells. Transcriptome levels agreed poorly with proteome changes overall. However, after filtering out low confidence microarray measurements, messenger RNA changes and SILAC ratios correlated very well for pheromone pathway components. Systems-wide, precise quantification directly at the protein level opens up new perspectives in post-genomics and systems biology.

Although the paper focuses on the larger amount of cell-wall proteins and proteins involved in pheromone response in haploid cells, the supplementary tables reveal similar biases for many other functional classes, including nucleosomes and cyclin-dependent kinase inhibitors. As many of these proteins are regulated during the cell cycle, I suspected that cell-cycle-regulated proteins might be more abundant in haploid cells relative to diploid cells.

To test this hypothesis, I divided the proteins quantified by the Mann group into two classes: dynamic proteins, which are encoded by genes that are periodically expressed during the cell cycle, and static proteins, which are encoded by genes that are expressed at a constant level (de Lichtenberg et al., 2005). For each class, I plotted the log₂-ratios of the protein levels in haploid and diploid cells:

The plot reeals a quite strong shift of dynamic proteins toward higher log-ratios; this difference is highly significant according to the Mann-Whitney U test (P < 10^-12). Proteins encoded by cell-cycle-regulated genes are thus in general more abundant in haploid budding yeast cells than in diploid cells.

Full disclosure: I currently collaborate with Matthias Mann and members of his group, and we will soon be colleagues a the Novo Nordisk Foundation Center for Protein Research.

Cite this post

Daub and coworkers from Matthias Mann’s group recently published a paper in Molecular Cell, describing a phosphoproteomics study of kinases during S and M phase of the mitotic cell cycle:

Kinase-selective enrichment enables quantitative phosphoproteomics of the kinome across the cell cycle.

Protein kinases are pivotal regulators of cell signaling that modulate each other’s functions and activities through site-specific phosphorylation events. These key regulatory modifications have not been studied comprehensively, because low cellular abundance of kinases has resulted in their underrepresentation in previous phosphoproteome studies. Here, we combine kinase-selective affinity purification with quantitative mass spectrometry to analyze the cell-cycle regulation of protein kinases. This proteomics approach enabled us to quantify 219 protein kinases from S and M phase-arrested human cancer cells. We identified more than 1000 phosphorylation sites on protein kinases. Intriguingly, half of all kinase phosphopeptides were upregulated in mitosis. Our data reveal numerous unknown M phase-induced phosphorylation sites on kinases with established mitotic functions. We also find potential phosphorylation networks involving many protein kinases not previously implicated in mitotic progression. These results provide a vastly extended knowledge base for functional studies on kinases and their regulation through site-specific phosphorylation.

In the study, they identified phosphorylation sites for 219 protein kinases, of which 159 showed differential phosphorylation (at least two-fold induction for at least one site) in S and/or M phase.

My collaborators at CBS and I have previously shown that transcriptional and posttranslational regulation (for example, phosphorylation by cyclin-dependent kinases) tend to target the same proteins (de Lichtenberg et al., 2005; Jensen et al., 2006). One should thus expect that the differentially regulated kinases have a tendency to be encoded by periodically expressed genes.

To test this hypothesis, I compared the phosphoproteomics data of Daub et al. to the cell-cycle microarray expression study by Whitfield et al. (2002). I was able to map 132 of the 159 kinases to the microarrays and found that 17 of them are encoded by the top-600 cycling genes. This corresponds to a significant (P < 0.001) two-fold overrepresentation of transcriptional cell-cycle regulation among the genes encoding kinases that are differentially phosphorylated during S and/or M phase.

One could imagine that this trend is not specific to kinases that are differentially phosphorylated during the cell cycle, but that it instead applies to kinases in general. To test this, I also mapped the 60 non-modulated kinases found by Daub et al. to the microarrays (Whitfield et al., 2002). Of the 54 kinases that could be mapped, only 3 are encoded by periodically expressed genes, which is almost exactly what is expected by random chance.

I next examined if timing of phosphorylation correlates with the timing of expression of the 17 kinases mentioned above. The kinases can be divided into three classes: phosphorylated in S phase, phosphorylated in M phase, and phosphorylated in both S and M phase. Notably, 13 of the 17 kinases fall in to the M phase class. Looking at the peak times of expression for these (that is when in the cell-cycle the corresponding mRNAs are most highly expressed) reveals that 8 of the 13 kinases are presumably synthesized in M phase only shortly before they become phosphorylated.

In summary, comparison of the phosphoproteomics data from Daub et al. (2008) and the microarray expression data from Whitfield et al. (2002) supports the view that transcriptional and posttranslational regulation tend to target the same proteins during the mitotic cell cycle. Moreover, it shows that for most of the kinases that are subject to such dual cell-cycle control, both expression and phosphorylation takes place during M phase when the cyclin-dependent kinase activity is maximal.

Full disclosure: I currently collaborate with Matthias Mann and members of his group, and we will soon be colleagues a the Novo Nordisk Foundation Center for Protein Research.

Cite this post

In 2001, Jeong and coworkers published a paper in Nature in which they showed that the central proteins in interaction networks, that is the proteins with the highest connectivity, are enriched for essential proteins. This publication has been highly influential as evidenced by the numerous subsequent publications on the importance of “hub” proteins. Several hypothesis have been published that try to explain why hubs are essential, for example that certain protein interactions are essential and that a protein with many interactions is thus more likely to be involved in at least one essential interaction (He and Zhang, 2006).

Yesterday, Zotenko and coworkers published a paper in PLoS Computational Biology in which they take a closer look at the cause of this phenomenon:

Why Do Hubs in the Yeast Protein Interaction Network Tend To Be Essential: Reexamining the Connection between the Network Topology and Essentiality.

The centrality-lethality rule, which notes that high-degree nodes in a protein interaction network tend to correspond to proteins that are essential, suggests that the topological prominence of a protein in a protein interaction network may be a good predictor of its biological importance. Even though the correlation between degree and essentiality was confirmed by many independent studies, the reason for this correlation remains illusive. Several hypotheses about putative connections between essentiality of hubs and the topology of protein-protein interaction networks have been proposed, but as we demonstrate, these explanations are not supported by the properties of protein interaction networks. To identify the main topological determinant of essentiality and to provide a biological explanation for the connection between the network topology and essentiality, we performed a rigorous analysis of six variants of the genomewide protein interaction network for Saccharomyces cerevisiae obtained using different techniques. We demonstrated that the majority of hubs are essential due to their involvement in Essential Complex Biological Modules, a group of densely connected proteins with shared biological function that are enriched in essential proteins. Moreover, we rejected two previously proposed explanations for the centrality-lethality rule, one relating the essentiality of hubs to their role in the overall network connectivity and another relying on the recently published essential protein interactions model.

What Zotenko et al. show is, in other words, that essential hubs tend to be highly connected with each other and hence form large “Essential Complex Biological Modules”. Table 7 in their paper lists the Gene Ontology terms associated with these modules; among the recurring themes are “rRNA metabolic process”, “mRNA metabolic process”, “RNA splicing”, “ribosome biogenesis and assembly”, and “proteolysis”. These Gene Ontology terms obviously correspond to well known protein complexes, namely the RNA polymerases, the spliceosome, the ribosome, and the proteoasome. The analysis of Zotenko et al. thus suggests that the much debated correlation between centrality and essentiality is simply a consequence of the fact that many of the large protein complexes in a eukaryotic cell are essential, which is hardly surprising considering that they have been conserved through more than two billion years of evolution (Brocks et al., 1999).

Edit: For more views on the results of Zotenko et al. see the discussion on FriendFeed.

Cite this post

I am now at the ISMB conference from where I will attempt to provide live coverage of the events. To avoid flooding this blog with posts related to the conference, I have set up a separate blog on Tumblr for this purpose. All my posts there will also appear on my FriendFeed.

This week Nature published a News piece by Declan Butler with the rather provocative title “PLoS stays afloat with bulk publishing”. Unsurprisingly, this caused a backlash from open-access advocates in general and science bloggers in particular. Jonathan Eisen posted the ironic response “Only Nature could turn the success of PLoS One into a model of failure”. For an overview of the many other responses from the blogosphere see the summary by Coturnix and the long debate on FriendFeed.

The core of the criticism by Declan Butler was directed against the business model of the Public Library of Science (PLoS), in particular that a large part of their total income is produced by “bulk publishing” in the “database” PLoS ONE with only “light” peer review. There is no point in denying that PLoS ONE is a major source of income for PLoS, that it publishes many papers, and that it is not a top-tier journal. Still, it is in my view an unnecessary provocation to refer to a journal from a competitor as a “database” and between the lines suggest that they do not perform proper peer review.

I have nothing against Nature Publishing Group (NPG) – they are in my view one of the more progressive publishers with initiative such as Connotea and Nature Network. However, I find the criticism by Declan Butler somewhat unfair, especially considering that NPG also has a considerable number of lower impact journals in their portfolio in addition to their lineup of Nature journals. To illustrate this point, I looked up the impact factors for all the PLoS and NPG journals that I could find (6 and 68, respectively) and plotted the distributions:

The average impact factors of the two publishers are remarkably similar 9.19 for PLoS and 9.39 for NPG, but the underlying distributions are very different. Notably, the high average impact factor of NPG’s journals is due to a fairly small number of journals with impact factors over 20, which are sufficient to offset the large number of journals with impact factors below 5. Consequently, the median impact factors are 9.03 for PLoS and only 4.88 for NPG.

I want to be the first to point out the caveats of this analysis. First, the analysis above did not take into account that each journal does not publish the same number of papers. However, weighting the journals by number of papers when calculating average impact factors shifts the balance in favor of PLoS (9.79 for PLoS vs. 9.46 for NPG). Second, the journal PLoS ONE does not have an impact factor yet and was thus not included in my analysis. Third, the criticism by Declan Butler was mainly targeting the fact that much of PLoS’ revenue is due to PLoS ONE. However, until NPG chooses to make available detailed financial reports like PLoS does, it is impossible to tell how much of their revenue comes from lower-impact journals.

That being said, the business models of PLoS and NPG do not look all that different based on bibliographic metrics alone.

Full disclosure: I am an associate editor of PLoS Computational Biology.

Cite this post

For use in presentations on literature mining, I did a back-of-the-envelope calculation of how much time I would be able to spend on each new biomedical paper that is published. Assuming that all papers were indexed in PubMed (which they are not) and that I could read papers 24 hours per day all year around (which I cannot), the result is that I could allocate approximately 50 seconds per paper. This nicely illustrates the point that no one can keep up with the complete biomedical literature.

When I discovered Wordle, which can turn any text into a beautiful word cloud, I thus wondered if this visualization method would be useful for summarizing a complete paper as a single figure. To test this, I extracted the complete text of three papers that I coauthored in the NAR database issue 2008. Submitting these to Wordle resulted in the three figures below (click for larger versions):

All in all, I think that Wordle does a pretty good job at capturing the essence of each paper: the first cloud shows that STITCH is a database of interactions between proteins and chemicals, the second cloud shows that NetworKIN is a database predictions related to the kinases and phosphorylation, and the third cloud shows that Cyclebase.org is a database of experiments on gene expression during the cell cycle. However, a paper describing a database might be easier to summarize that a typical research paper.

As a final test, I therefore submitted the complete text from my paper “Evolution of Cell Cycle Control – Same molecular machines, different regulation”, which describes the somewhat complex concept of just-in-time assembly to Wordle (click for larger version):

The result is rather less impressive than for the papers from the NAR database issue. Although the word cloud does contain a good selection of words, it fails to convey the main message. I think a large part of the problem is the splitting of multiwords; for example, “cell cycle” becomes two separate terms “cell” and “cycle”. Another problem is that words from different sections of the paper are mixed, which blurs the messages. These two issues could be solved by 1) detecting multiwords and considering them as single tokens, and 2) sorting the terms according to where in the paper they are mainly used.

Cite this post

I yesterday blogged about how the protein half-life data from the O’Shea lab fit well with my earlier analyses of transcriptional regulation during the budding yeast cell cycle and with the just-in-time assembly hypothesis. However, I have now realized that the same data set can be used to test the validity of the sequence-based predictions of protein degradation signals that I relied on for the cell-cycle study.

To this end, I divided the budding yeast proteome into six groups: proteins with a D-box, proteins without a D-box, proteins with a KEN-box, proteins without a KEN-box, proteins with a PEST region, and proteins without a PEST region. For each of these six groups of proteins, I simply plotted the distribution of protein half-lives as a histogram:

The figure shows that for all three degradation signals, proteins with the sequence motif tend to have shorter half-lives than proteins without the motif. These differences are all statistically significant according to the Mann-Whitney U test (D-box, P < 10^-6; KEN-box, P < 0.02; PEST region, P < 10^-15). It is noteworthy that the KEN-box motif gives a far weaker correlation with protein half-live than the two other degradation signals, as it was also the only degradation signal that did not correlate with transcriptional cell-cycle regulation in budding yeast (see supplementary information of Jensen et al., 2006).

In summary, proteins that contain putative degradation signals have significantly shorter half-lives than proteins that do not contain such signals. The only caveat is that long sequences are more likely to match the sequence motifs, and that O’Shea and colleagues found a negative correlation between sequence length and protein half-life. The correlations described here could thus be a secondary effect; however, it is also possible that the presence of degradation signals in long sequences is the missing explanation for their short half-lives.

Cite this post

In relation to an entirely different analysis than the one I will describe here, I downloaded the protein half-life data for budding yeast that was published in PNAS by the O’Shea lab about two years ago:

Quantification of protein half-lives in the budding yeast proteome

A complete description of protein metabolism requires knowledge of the rates of protein production and destruction within cells. Using an epitope-tagged strain collection, we measured the half-life of >3,750 proteins in the yeast proteome after inhibition of translation. By integrating our data with previous measurements of protein and mRNA abundance and translation rate, we provide evidence that many proteins partition into one of two regimes for protein metabolism: one optimized for efficient production or a second optimized for regulatory efficiency. Incorporation of protein half-life information into a simple quantitative model for protein production improves our ability to predict steady-state protein abundance values. Analysis of a simple dynamic protein production model reveals a remarkable correlation between transcriptional regulation and protein half-life within some groups of coregulated genes, suggesting that cells coordinate these two processes to achieve uniform effects on protein abundances. Our experimental data and theoretical analysis underscore the importance of an integrative approach to the complex interplay between protein degradation, transcriptional regulation, and other determinants of protein metabolism.

The idea that transcriptional regulation goes hand-in-hand with protein degradation is fully consistent with the just-in-time assembly hypothesis. I thus examined the distributions of protein half-lives for dynamic (i.e. periodically expressed) and static (i.e. not periodically expressed) proteins:

The histogram suggests that dynamic proteins are shifted towards shorter half-lives relative to static proteins. The difference is indeed statistically significant according to the Mann-Whitney U test (P < 10^-4). This result supports the sequence-based observation that dynamic proteins contain more D-box, KEN-box, and PEST degradation signals than static proteins.

I next tested if the half-life of the dynamic proteins varies during the cell cycle by make scatter plot of the protein half-life as function of the time of peak expression for the corresponding mRNA:

There appears to be no correlation. Together, these analyses indicate that dynamic proteins have shorter half-lives than static proteins, irrespective of when in the cell cycle they are expressed.

Cite this post

Over the years several microarray time-course experiments have been performed to identify the genes that are transcriptionally regulated during the mitotic cell cycle, i.e the periodically expressed genes. Moreover, bioinformaticians have developed many different computational methods for identifying the periodically expressed genes from microarray time-course data.

Below is a list of the experimental and computational analyses of the budding yeast cell cycle that I am aware of (please notify me if you know of other microarray experiments or computational methods):

1. Cho et al., Mol. Cell, 1998
2. Spellman et al., Mol. Biol. Cell, 1998
3. Zhao et al., Proc. Natl. Acad. Sci. USA, 2001
4. Langmead et al., Proc. IEEE Comput. Soc. Bioinformatics Conf., 2002
5. Langmead et al.,RECOMB, 2002
6. Langmead et al., J. Comput. Biol., 2003
7. de Lichtenberg et al., J. Mol. Biol., 2003
8. Johansson et al., Bioinformatics, 2003
9. Wichert et al., Bioinformatics, 2004
10. Lu et al., Nucleic Acids Res., 2004
11. Luan and Li, Bioinformatics, 2004
12. de Lichtenberg et al., Bioinformatics, 2005
13. de Lichtenberg et al., Yeast, 2005
14. Willbrand et al., Bioinformatics, 2005
15. Ahdesmäki et al., BMC Bioinformatics, 2005
16. Chen, BMC Bioinformatics, 2005
17. Qiu et al., Conf. Proc. IEEE Eng. Med. Biol. Soc., 2005
18. Qiu et al., Bioinformatics, 2006
19. Andersson et al., BMC Bioinformatics, 2006
20. Gan et al., Int. Conf. Pattern Recog., 2006
21. Glynn et al., Bioinformatics, 2006
22. Ahnert et al., Bioinformatics, 2006
23. Lu et al., Bioinformatics, 2006
24. Xu et al., LSS Comput. Syst. Bioinformatics Conf., 2006
25. Pramilla et al., Genes Dev., 2006
26. Liew et al, BMC Bioinformatics, 2007
27. Lu et al., Genome Biol., 2007
28. Morton et al., Stat. Appl. Genet. Mol. Biol., 2007
29. Rowicka et al., Proc. Natl. Acad. Sci. USA, 2007
30. Gauthier et al., Nucleic Acids Res., 2008
31. Orlando et al., Nature, 2008

These studies have reported a mixture of ranked and unranked lists of periodically expressed genes. By that I mean that some studies provided a list of genes sorted according to how periodic the expression profiles appear, whereas others simply provide a list of the genes deemed periodic. For the ranked lists, I first checked the publications to see if the authors suggested a cutoff for the number of periodically expressed genes, in which case I followed their recommendations. If the authors suggested multiple lists of varying confidence, I used the highest-confidence list. If no cutoff was proposed, I selected the top-300 genes if the list was based on a single time course and the top-500 genes if the list was based on three or more time courses. It should be noted that both of these cutoffs are on the conservative side since most studies propose 800 or more periodically expressed genes when combining multiple expression time courses.

This meta-analysis resulted in a list of more than 4200 budding yeast genes that are periodically expressed according to at least one of the methods listed above; that is more than two-thirds of all genes encoded by the budding yeast genome!

To investigate further how such a large number of genes can have been proposed to be periodically expressed, I plotted how many of these genes are on how many of the lists of periodically expressed genes:

The histogram reveals that the majority of the over 4200 genes have been proposed by only one or two analyses. It seems reasonable to assume that the genes that have been proposed as periodically expressed by only one or a few methods are less likely to be correct than the ones that many methods agree on. Also, one could expect that taking the consensus of many methods would yield a more reliable answer than using just a single method.

To test these two hypotheses, I compared two different ways of identifying the periodically expressed genes:

1. Ranking the genes based on a single scoring scheme that combines all the available experimental data (Gauthier et al., Nucleic Acids Res., 2008)
2. Ranking the genes based on vote among 30 different methods (not 31; the analysis by Orlando and coworkers was left out of the voting as this dataset is not included in Cyclebase.org)

To benchmark the two methods, I compared the ranked lists to a set of target genes for cell-cycle transcrition factors identified in genome-wide ChIP-on-chip experiments and plotted the fraction of these that were identified as function of the number of genes proposed to be periodically expressed:

The plot confirms that genes proposed to be periodically by multiple methods are more likely to be targets of cell-cycle transcription factors, and are hence more likely to truly be subject to transcriptional cell-cycle regulation. However, it also shows that the list obtained by voting among 30 methods is a bit worse than what is obtained by using the single best method.

This result may come as a surprise to many since meta-servers that combine multiple prediction methods have in the past proven very successful for many other bioinformatics tasks. I suspect that the approach fails in this case for two reasons: first, many of the analyses included perform considerably worse than the best one, and second, most of the methods make use of only half of the available experimental data. It may thus be possible to obtain better results by selecting only a subset of the methods and rerunning each of them on all the available data. So far, however, dictatorship seems to work better than democracy for identification of periodically expressed genes.

Cite this post

I recently took a look at colonization of titles and found that the fraction of papers with colons in their titles is increasing steadily. Intuitively, one would thus expect that the average length of the titles has also increased. The plot below shows that this is indeed the case (not that the y-axis does not begin at zero):

The average title length has increased from 8.5 words in 1950 to 12.5 words in 2008. Strangely, the increase is almost perfectly linear except for a fluctuation in the early 60s – I have no idea why this is the case.

But is the title length of a paper important? I personally expected that papers with short, catchy titles would be cited more than papers with longer, more complex titles. Lacking citation information for individual publications, I thus calculated average title length for publications from each journal and correlated it with the ISI impact factor of the corresponding journal:

No correlation is observed between the impact factor of a journal and the average title length of the papers published therein. So we can conclude that – at least for titles of scientific papers – size does not matter.

Cite this post

You have probably noticed that a high fraction of scientific papers have colons in their titles. Several people have written humorous commentaries on this. Although these authors clearly see the use of colons as a growing trend, they did not present hard evidence for the increase in the usage of colons in the titles of scientific publications.

Out of curiosity, I thus wrote a small script to count the fraction of papers in Medline that have colons in their titles for each of the past 25 years. The result is shown in the plot below (note that the y-axis does not start at zero):

The conclusion is very clear: the fraction of titles with colons has increased linearly from 15% to 24% over the past 20 years. One could object that this effect may be explained by the increase in apologies (which often have a title “Retraction: …”) or by the NAR special issues on databases and web servers (which contain hundreds papers with titles such as “YADB: yet another database”). However, these add up to less than 2% of the papers with colonized titles and are thus insufficient to explain the observed 9% increase.

Cite this post

I have long used a data integration approach to obtain a global picture of eukaryotic cell-cycle regulation. The cell cycle is a popular research topic in part because of its importance for cancer research. I thus recently compared microarray expression data on the human cell cycle to genes with mutations that have been causally implicated in various forms of cancer.

From the Cancer Genome Project website, I downloaded a list of 353 human genes that are implicated in cancer. Using the identifier mapping file from STRING, I was able to automatically map 338 of these genes to the set of human genes from Ensembl that I used in earlier cell-cycle studies. 295 of the 338 genes were present on the microarrays used in the cell-cycle expression study by Whitfield et al. (2002). However, only 23 of these are among the 600 periodically expressed genes identified in the reanalysis by Jensen et al. (2006). The many numbers are illustrated in the diagram below:

By random chance, 295*600/12097 = 15 of the 295 genes would be expected to be periodically expressed, and the enrichment is thus only a bit over 1.5 fold. Although this enrichment is statistically significantly (P < 3%, Fisher’s exact test), the correlation is clearly not strong enough to allow prediction of novel cancer genes.

My step was to look at the evolutionary conservation of the 23 periodically expressed cancer genes. Only 12 of them belong to an orthologous group. Half of them do thus not appear to have orthologs in budding yeast, fission yeast, or Arabidopsis thaliana. Only three periodically expressed cancer genes have orthologs in all of these organisms. One of these genes is periodically expressed onlt in human, one in human and fission yeast, and one in all four organisms (a histone subunit).

In summary, it seems that one cannot say much about cancer based on cell-cycle mRNA expression data. This is perhaps not surprising considering that the transcriptional regulation does not seem to vary much between cancer cells and normal cells.

Cite this post

The groups of Ziv Bar-Joseph and Itamar Simon recently published a paper in PNAS on a new microarray study of the cell cycle of primary human fibroblasts:

Genome-wide transcriptional analysis of the human cell cycle identifies genes differentially regulated in normal and cancer cells

Characterization of the transcriptional regulatory network of the normal cell cycle is essential for understanding the perturbations that lead to cancer. However, the complete set of cycling genes in primary cells has not yet been identified. Here, we report the results of genome-wide expression profiling experiments on synchronized primary human foreskin fibroblasts across the cell cycle. Using a combined experimental and computational approach to deconvolve measured expression values into ‘‘single-cell’’ expression profiles, we were able to overcome the limitations inherent in synchronizing nontransformed mammalian cells. This allowed us to identify 480 periodically expressed genes in primary human foreskin fibroblasts. Analysis of the reconstructed primary cell profiles and comparison with published expression datasets from synchronized transformed cells reveals a large number of genes that cycle exclusively in primary cells. This conclusion was supported by both bioinformatic analysis and experiments performed on other cell types. We suggest that this approach will help pinpoint genetic elements contributing to normal cell growth and cellular transformation.

In contrast to the earlier study by Whitfield et al. (2002), which was performed on HeLa cells, Ziv Bar-Joseph et al. worked on non-transformed fibroblasts. The dataset thus offers a first global view of the differences between the cell cycle of normal human cells and that of cancer cells.

To compare their list of cell-cycle-regulated human genes to the one the I have used so far, I mapped their 480 genes to Ensembl using the mapping file from the STRING database. This resulted in a list of 410 genes, that is 70 genes could not be mapped by the automatic procedure. Whereas this is far from a perfect mapping, it is sufficient to judge the quality of the list.

The plots below show the fraction of a benchmark set that is identified as function of the number of genes that is proposed to be periodically expressed during the cell cycle. In each plot, I compare the results for the list of 410 obtained from the new study by Bar-Joseph et al., the analysis by Whitfield et al., and the reanalysis of the latter dataset by Jensen et al. (2006) (available from Cyclebase.org). To make the comparison as fair as possible, I only considered the subset of genes that were present in both microarray designs. The first plot uses as benchmark a set of 63 genes that have been identified as periodically expressed in targeted small-scale studies:

I also benchmarked the three gene lists against a second benchmark set, which consists of predicted target genes of E2F cell-cycle transcription factors:

Both benchmarks suggest that the three lists are of very comparable quality, but that the list by Whitfield and coworkers is much more inclusive than the one from Bar-Joseph and coworkers. In other words, the former list has better sensitivity whereas the latter has better specificity. This is consistent with the results presented by Bar-Joseph et al., who conclude that their list is more reliable than the previously published list. However, this is probably not due to better quality of the raw expression data, since reanalysis of the data by Whitfield et al. yielded a list with almost identical sensitivity and specificity (that is the red curve is very close to the blue cross in both plots).

Although the two lists of periodically expressed are of comparable quality, they may still contain very different sets of genes. I therefore decided to compare the list of genes that are periodically expressed in the time course on primary fibroblasts and in each of the four time courses on HeLa cells. To make this comparison as easy as possible, I selected the top-364 cycling genes from each of the four HeLa time courses based on the reanalysis by Jensen et al. (2006). The ten Venn diagrams below show all pairwise comparisons of the five lists of 364 genes each:

The average overlap between the list by Bar-Joseph et al. and an experiment from Whitfield et al. is 114 genes. By comparison, the average overlap between the top-364 lists from two individual experiments from Whitfield et al. is 123 genes. Although the overlap may seem low, I thus believe that it is due to the poor reproducibility between microarray time courses rather than due to genuine differences between primary fibroblasts and HeLa cells as suggested by Bar-Joseph and colleagues.

Although cancer cells have to circumvent the regulatory mechanisms that would normally prevent cell proliferation, the cell cycle itself appears to function the same way as in normal cells. In other words, the difference does not lie in the “engine” but in the “brakes”, which have been sabotaged in cancer cells.

Cite this post

Atsushi Miyawaki’s lab from RIKEN has recently published a Cell paper that describes a novel approach for how to monitor cell-cycle progression of individual cells:

Visualizing spatiotemporal dynamics of multicellular cell-cycle progression

The cell-cycle transition from G1 to S phase has been difficult to visualize. We have harnessed antiphase oscillating proteins that mark cell-cycle transitions in order to develop genetically encoded fluorescent probes for this purpose. These probes effectively label individual G1 phase nuclei red and those in S/G2/M phases green. We were able to generate cultured cells and transgenic mice constitutively expressing the cell-cycle probes, in which every cell nucleus exhibits either red or green fluorescence. We performed time-lapse imaging to explore the spatiotemporal patterns of cell-cycle dynamics during the epithelial-mesenchymal transition of cultured cells, the migration and differentiation of neural progenitors in brain slices, and the development of tumors across blood vessels in live mice. These mice and cell lines will serve as model systems permitting unprecedented spatial and temporal resolution to help us better understand how the cell cycle is coordinated with various biological events.

The clever idea was to fuse a red- and a green-emitting fluorescent protein to Cdt1 and Geminin, respectively. Cdt1 is ubiquitinated by SCF^Skp2 at the onset of S phase, which causes it to be rapidly degraded by the proteasome, whereas Geminin is targeted for proteolytic degradation by APC^Cdh1 in late M phase. By fluorescent labeling of two proteins, Miyawaki and colleagues managed to make mouse cells that become increasingly red during G1 phase, yellow around the G1/S transition, and increasingly green through S, G2, and M phase. It is thus possible to monitor the cell-cycle states of individual cells with a microscope.

The movie below follows a few HeLa cells for 3-4 cell cycles:

The authors also show how their construct can be used for imaging the cell-cycle state of the cells in a slice of a mouse brain or a mouse embryo. I expect that this will become an indispensable tool for unraveling the links between cell-cycle control and developmental processes.

For more details, I strongly recommend that you read Jake Young’s post at Pure Pedantry.

Cite this post

There seems to be a new trend in computational biology: worrying about sequence alignments. Over the past couple of months, two high-profile papers have appeared that flaws related to sequence alignment methods.

The first paper appeared in Science Magazine in January this year. Wong and coworkers describe how uncertainties in multiple alignments can lead to errors in different phylogenetic trees:

Alignment Uncertainty and Genomic Analysis

The statistical methods applied to the analysis of genomic data do not account for uncertainty in the sequence alignment. Indeed, the alignment is treated as an observation, and all of the subsequent inferences depend on the alignment being correct. This may not have been too problematic for many phylogenetic studies, in which the gene is carefully chosen for, among other things, ease of alignment. However, in a comparative genomics study, the same statistical methods are applied repeatedly on thousands of genes, many of which will be difficult to align. Using genomic data from seven yeast species, we show that uncertainty in the alignment can lead to several problems, including different alignment methods resulting in different conclusions.

The second paper appeared in Nature Biotechnology. Styczynski and coworkers discovered that the most commonly used substitution matrix, BLOSUM62, was calculated wrongly:

BLOSUM62 miscalculations improve search performance

The BLOSUM family of substitution matrices, and particularly BLOSUM62, is the de facto standard in protein database searches and sequence alignments. In the course of analyzing the evolution of the Blocks database, we noticed errors in the software source code used to create the initial BLOSUM family of matrices (available online). The result of these errors is that the BLOSUM matrices — BLOSUM62, BLOSUM50, etc. — are quite different from the matrices that should have been calculated using the algorithm described by Henikoff and Henikoff. Obviously, minor errors in research, and particularly in software source code, are quite common. This case is noteworthy for three reasons: first, the BLOSUM matrices are ubiquitous in computational biology; second, these errors have gone unnoticed for 15 years; and third, the ‘incorrect’ matrices perform better than the ‘intended’ matrices.

Upon casual reading of these publications, one could get the idea that over a decade of work based on alignments, sequence similarity searches, and molecular evolution is wrong. Fortunately, this does not appear to be the case.

Starting with the second paper, I applaud the authors for discovering a mistake in such an established method, and I agree with them that it is remarkable that it has not been noticed before. However, I do not think that it is surprising that the ‘incorrect’ matrices work very well. Although they were not calculated as intended, the BLOSUM matrices have become the de facto standard precisely because they work as well as they do.

Regarding the first paper, I think it is fair to say that anyone working on multiple alignments and phylogeny are well aware that uncertain alignments can lead to wrong phylogenetic trees. This is why almost everyone uses programs like Gblocks to remove the ambiguous parts of their alignments before moving on to constructing phylogenetic trees. Unfortunately, Wong et al. instead constructed two sets of trees for each of the six multiple alignment methods: one based on the complete alignments, and one in which they excluded all gapped sites from the phylogenetic analysis. The latter is not equivalent to using a blocked alignment, since not all ambiguously aligned sites contain gaps, and since not all sites with gaps are ambiguously aligned.

Wong and coworkers subsequently compared the trees that they obtained using the six different alignment programs and found disagreements for almost half of all yeast proteins. This number may sound shockingly high, but I find it to be misleading in several ways. First, “disagreement” was defined as at least one of the six trees disagreeing with the others – much of the disagreement could thus be due to a single poorly performing alignment program. This definition also implies that the results can only get worse by adding more alignment methods to the comparison. Second, the comparison was not limited to the trees that are supported by bootstrap analysis – much of the disagreement is thus due to trees that we already know should not be trusted.

In my view, it would be more fair to make the comparison along the following lines:

• Align the sequences as done by Wong et al.
• Remove ambiguously aligned sites with Gblocks
• Construct phylogenetic trees based on the blocked alignments
• Calculate the bootstrap support for each tree
• Discard trees with poor bootstrap support
• Calculate the agreement on tree topology for each pair of alignment methods

This procedure will ensure that trees are not distorted by the unreliable parts of the alignments, that comparisons are not based on trees we know are unreliable, that the results are not skewed by a single poorly performing alignment method, and that the numbers remain comparable if more alignment methods are added. I have already downloaded all the alignments and run then through Gblocks; please let me know if you would like to continue the analysis from that step, and I will arrange a way to transfer the files.

Time might prove me wrong, but I expect that such an analysis will show that alignment uncertainty is not a major factor that needs to be taken into account when constructing phylogenetic trees.

Cite this post

I am now back from two weeks in Italy where I experimented with live blogging. You have probably noticed that some presentations from the meeting at CoSBi in Trento were covered on Buried Treasure within a matter of minutes of them ending. Also, quite a number of pictures were posted in the associated Picasa web album while the presentations were still ongoing. Here is a brief explanation of how I planned to pull this off and how it worked in practice.

My original plan was to use WordPress through the web browser on my smartphone together with a foldable bluetooth keyboard. This was how I first imagined that my live-blogging platform would look:

It seemed a good idea at the time, but there were a couple of “minor” problems:

• The mobile web interface for WordPress does not enable you to upload pictures.
• The full web interface for WordPress works neither in Microsoft Internet Explorer Mobile nor in Opera Mobile.

I thus started looking around for alternative clients for WordPress and eventually found ShoZu, which allows you to upload pictures from your phone to a variety of services including WordPress blogs and Picasa web albums. However, it is not a true blogging tool and only enables you to write a short description for each picture. I thus accepted to the real blog posts would be written on my laptop, whereas the following platform would be used for live blogging in the form of images with short descriptions:

This seemed like an even better idea at the time, but again I ran into a few technical problems:

• Due to strange combinations of firewalls, HTTP proxies, and complex login web pages, I never managed to get my smartphone reliably connected to the internet.
• The camera in the smartphone was unable to take even half decent picture under the poor light conditions.

In reality, I thus ended up using my old Apple PowerBook G4 and my Lumix TZ3 camera. They got the job done in terms of covering the presentations, but live blogging from poster sessions was practically impossible.

I have now put on my thinking cap to come up with a live-blogging platform that would work for poster sessions. You generally have too little time for too many posters, so it has to be very fast to snap a photo and post it. The light is often poor and people tend to use too small fonts on their posters, so you need a good camera to get a readable result. Finally, the lack of space and tables prevents you from using a laptop. The Eye-Fi Card  might be a solution as it would enable me to upload images directly from my camera to, for example, a Picasa web album or Flickr. Please let me know if you have any ideas, experiences, or thoughts on this.

The group of Donald F. Hunt at University of Virginia has recently published a paper in PNAS that describes a new phosphoproteomics study of budding yeast:

Analysis of phosphorylation sites on proteins from Saccharomyces cerevisiae by electron transfer dissociation (ETD) mass spectrometry

We present a strategy for the analysis of the yeast phosphoproteome that uses endo-Lys C as the proteolytic enzyme, immobilized metal affinity chromatography for phosphopeptide enrichment, a 90-min nanoflow-HPLC/electrospray-ionization MS/MS experiment for phosphopeptide fractionation and detection, gas phase ion/ion chemistry, electron transfer dissociation for peptide fragmentation, and the Open Mass Spectrometry Search Algorithm for phosphoprotein identification and assignment of phosphorylation sites. From a 30-microg (approximately 600 pmol) sample of total yeast protein, we identify 1,252 phosphorylation sites on 629 proteins. Identified phosphoproteins have expression levels that range from <50 to 1,200,000 copies per cell and are encoded by genes involved in a wide variety of cellular processes. We identify a consensus site that likely represents a motif for one or more uncharacterized kinases and show that yeast kinases, themselves, contain a disproportionately large number of phosphorylation sites. Detection of a pHis containing peptide from the yeast protein, Cdc10, suggests an unexpected role for histidine phosphorylation in septin biology. From diverse functional genomics data, we show that phosphoproteins have a higher number of interactions than an average protein and interact with each other more than with a random protein. They are also likely to be conserved across large evolutionary distances.

As is so often the case with experimental papers, no comparison is provided to earlier studies. I thus decided to compare the set of phosphoproteins identified by Hunt and coworkers to the set of Cdc28p substrates identified in two studies by the Morgan lab as well as to the proteome-wide, sequence-based predictions made by NetPhosK:

The Venn diagram obviously shows that each of the three sets contains a considerable number of phosphoproteins that are not present in any of the other sets. This was to be expected since the three methods are fundamentally very different. The dataset from the Hunt lab includes proteins that are phosphorylated by other kinases than Cdc28p; however, it is limited in the sense that low-abundance phosphopeptides are typically missed by MS studies. Conversely, the set from the Morgan lab consists only of Cdc28p substrates, but is likely to have much better coverage of low-abundance phosphoproteins. Finally, the set of Cdc28 substrates from NetPhosK is likely to contain a considerable number of false positives as they are predicted from the protein sequence alone.

As a matter of fact, I find the overlap between the three sets to be surprisingly good. Even if we assume that the dataset from the Morgan lab contains no false positives, the overlap suggests that the new dataset from Hunt and coworkers captures one third of all phosphoproteins in budding yeast; assuming errors in both datasets increases this estimate. It is also noteworthy that NetPhosK misses only 22% of the Cdc28p that were identified by the Morgan lab and supported by the new data from the Hunt lab, although this high coverage is probably obtained at the price of many false positive predictions.

Cite this post

You may have heard about the controversial movie “Expelled: No intelligence Allowed” by Ben Stein in which people behind the intelligent design movement whine about being suppressed the scientific community. The truth is obviously that intelligent design is not a falsifiable theory and hence simply does not qualify as science.

However, the movie is also controversial in other respects. To start with the producers conned both Richard Dawkins and fellow blogger PZ Myers into participating in the movie by interviewing them under false pretense.

Richard Dawkins and PZ Myers thus both registered for participating in a public screening of the movie. But while queuing up for the movie, PZ Myers was identified by security officers and told to leave the premises – immediately! Oh the irony, oh the double standard. They make a movie about suppression of views, they call it “Expelled”, and then they expel a person whom you conned into participating in the movie because you disagree with his views.

But it gets even better. The very same security officers were apparently oblivious to the fact that Richard Dawkins was standing right next to PZ Myers and thus let him enter to watch the movie. PZ Myers immediately wrote a blog post about it, while the movie was still being shown to the audience – including Richard Dawkins.

After the movie, Richard Dawkins (of course) stood up and asked why PZ Myers was not allowed to see the movie. The answer? Because he did not have a ticket and was thus a gate crasher! Very interesting explanation since it was not a ticket event – you simply had to register a seat, which PZ Myers had done.

The two gentlemen have now posted an interesting little discussion on YouTube in which they humorously describe the incident as well as just how bad the movie really is:

Richard Dawkins also reveals that Expelled includes one of the beautiful movies produced by the multimedia team at Harvard. You really have to wonder if they actually got permission for that, if they conned the people at Harvard as well, or if they just resorted to plain old plagiarism. In any case, this has to be one of the biggest PR disasters ever made by the intelligent design movement.

Expelled from Expelled: no intelligence involved.

I have now arrived in Bertinoro where I will be lecturing on the 8th Course in Bioinformatics for Molecular Biologists. And after a fight with network configuration and power outages, I also eventually managed to get online.

All the speakers are housed at the castle, which has a fantastic view over the surrounding area – also by night:

The scientific part of the meeting was kicked off by H. Werner Mewes:

I am sure there will be many interesting lectures to follow – and I hope that the audience will think that mine is one of them.

Orkun Soyer has just finished his excellent presentation at CoSBi on the use of toy models for understanding the principles that govern biological pathways, in particular signaling pathways. One can obviously imagine several scenarios for how pathways came about:

The key point, however, is that we might be able to understand something about pathways through computational studies of simple toy models. The toy model discussed throughout the talk was bacterial chemotaxis:

The idea is that evolution can to some extend be approximated as an optimization process, in which the objective function corresponds to fitness. In case of the “tumble or swim” problem, computational simulations allowed simple regulatory network to evolve that mimic the food-finding behavior of bacteria.

He also presented an interesting view on how biological complexity has evolved. The idea is to show how complex systems can evolve even if assuming a (weak) selection against complexity:

I think that his results provide a lot of insight into how real signaling may have evolved, although all the simulations are based on simplistic toy models. I recommend that you download Orkun Soyer’s slides if you want to know more.

This talk ends the Computational and Systems Biology course at CoSBi.

As most readers of this blog are probably aware, Mohammad Warda and Jin Han published a paper in Proteomics that contained several pages of text copied from unreferenced sources. Exactly one month ago it was thus retracted “due to a substantial overlap of the content of this article with previously published articles in other journals”.

Plagiarism is, however, not the main issue. The paper by Warda and Han also claimed to disprove the endosymbiotic origin of mitochondria, mentioned fingerprints of a mighty creator, and proposed mitochondria to be the missing link between the body and the preserved wisdom of the soul!

It remains a mystery how a manuscript with such unsubstantiated claims was accepted for publication in a respectable, peer-reviewed journal. The retraction notice by Proteomics made no attempt to explain this, and their approved draft press release merely states that it was due to “human error”. I would have been really worried if it could happen without human error being involved. Although this draft was approved a month ago, the final version is nowhere to be found on the internet, also not on the Proteomics website. I thus wonder if an official press release was ever published.

Attila Csordas, PZ Myers, Steven Salzberg, and I have decided to mark the one month anniversary of the retraction by pointing out the important questions that still remain to be answered by the Editor in Chief of Proteomics, Prof. Michael J. Dunn.

The manuscript contains four parts with unsupported claims that should have been caught by any peer reviewer or editor:

1. Title – “Mitochondria, the missing link between body and soul”.
2. Abstract – “These data are presented with novel proteomics evidence to disprove the endosymbiotic hypothesis of mitochondrial evolution that is replaced in this work by a more realistic alternative”.
3. Section 3.4 – “More logically, the points that show proteomics overlapping between different forms of life are more likely to be interpreted as a reflection of a single common fingerprint initiated by a mighty creator than relying on a single cell that is, in a doubtful way, surprisingly originating all other kinds of life”.
4. Conclusions – “We realize so far that the mitochondria could be the link between the body and this preserved wisdom of the soul devoted to guaranteeing life”.

My questions to Michael J. Dunn are when in the publication process these parts first appeared:

1. Were they already in the initial version that was submitted to Proteomics and sent out for peer review?
2. Did they appear in a revised version that was sent to the peer reviewers?
3. Were they introduced in a revised version that was accepted without sending it to the reviewers?
4. Or were they added at the copy editing stage, that is after the manuscript had formally been accepted?

I want to make explicit that the aim with these questions is not to place the blame but to elucidate what went wrong in the publication process. To prevent similar incidents inthe future, it is important to know whether the editor and the peer reviewers overlooked glaring flaws of the manuscript or if the flawed parts were introduced after peer review. It is not important who the editor and the peer reviewers are. I sincerely hope that Prof. Dunn will help improve the procedures for peer reviewed publication by answering the questions in this post and in the related posts on PIMM, Pharyngula, and Genomics, Evolution, and Pseudoscience.

Cite this post

Gabor Balazsi has just finished a very interesting presentation on the interplay between molecular networks, gene expression noise, and evolutionary selection – here is the opening slide:

In the first part of his talk he gave a nice introduction to global network topology and network motifs – this should be nothing new to people familiar with the work of the Barabasi and Alon labs. He also explained the “Commander, Intermediate, Executor” model for hierarchical regulatory networks, which I had personally not heard about before, and the concept of “origons”, which seems quite use for understanding the response of large signaling networks to environmental cues.

The second part of his talk was about stochastic noise in gene expression. Genetically identical cells in a culture may express the same protein at different levels; this is a result of random noise influencing transcription, mRNA degradation, translation, and protein degradation. This is simply a consequence of low copy numbers giving rise to stochastic, as opposed to deterministic, behavior.

Finally, he talked about how noise at the level of gene expression can influence the survival of species in a changing environment. This part of his talk was kicked off with the funniest slide of his presentation:

I guess it should be seen as a lesson on how not to do. He made some very good points about how noise plays hardly any role in multicellular organisms that reproduce sexually. By contrast, stochastic variation within clonal bacterial cultures provides much higher chance of survival when faced with sudden stress such treatment with anti-bacterial drugs. I would have liked to hear more about this, but unfortunately there was not much time left for this part of the presentation due to technical problems with the projectors. It looks like Guy Shinar picked the safe strategy for his presentation.

All in all, I found it to be a really inspiring talk. I have uploaded his slides in case if you want to take a look at it.

Fifteen minutes ago, Attila Csikasz-Nagy opened the Computational and Systems Biology Course at CoSBi in Trento, Italy:

Over the coming week, I will be covering the most interesting presentations and posters here on the blog and in the Picasa web album.

David Botstein’s group at Princeton recently published a paper in Molecular Biology of the Cell with the title “Coordination of Growth Rate, Cell Cycle, Stress Response, and Metabolic Activity in Yeast”. As described in their abstract, they found interesting several correlations between the transcriptional responses to changes in growth rate and the regulation in response to stress and during the metabolic cycle:

We studied the relationship between growth rate and genome-wide gene expression, cell cycle progression, and glucose metabolism in 36 steady-state continuous cultures limited by one of six different nutrients (glucose, ammonium, sulfate, phosphate, uracil, or leucine). The expression of more than one quarter of all yeast genes is linearly correlated with growth rate, independent of the limiting nutrient. The subset of negatively growth-correlated genes is most enriched for peroxisomal functions, whereas positively correlated genes mainly encode ribosomal functions. Many (not all) genes associated with stress response are strongly correlated with growth rate, as are genes that are periodically expressed under conditions of metabolic cycling. We confirmed a linear relationship between growth rate and the fraction of the cell population in the G0/G1 cell cycle phase, independent of limiting nutrient. Cultures limited by auxotrophic requirements wasted excess glucose, whereas those limited on phosphate, sulfate, or ammonia did not; this phenomenon (reminiscent of the “Warburg effect” in cancer cells) was confirmed in batch cultures. Using an aggregate of gene expression values, we predict (in both continuous and batch cultures) an “instantaneous growth rate”. This concept is useful in interpreting the system-level connections among growth rate, metabolism, stress, and the cell cycle.

Because of my interest in cell cycle, their results regarding growth rate and cell-cycle regulation caught my attention. In Figure 6 of their paper, Brauer et al. show the slope distribution for the genes belonging to each of the phase-specific clusters defined by Spellman et al. (1998). The only trend they observe is that genes expressed at the G1/M transition.

I decided to redo the cell-cycle part of their analysis in a slightly different manner, hoping that I would be able to get a stronger signal than they did. Rather than using the 800 periodically expressed genes proposed by Spellman et al. (1998), I thus made use of the list of 600 periodically expressed genes from de Lichtenberg et al. (2005). Like Brauer et al., I found no difference in growth-rate response between cell-cycle-regulated genes and other genes. To analyze the phase-specific expression, I chose to plot the peak time distributions for genes that are up- and down-regulated in response to increasing growth rate:

In agreement with Brauer et al., genes that are down-regulated at high growth rates appear to have a striking preference for being expressed at the G1/M transition. However, manual inspection of these genes revealed that more than half of them belong to the Y’ family of DNA helicases, which are encoded by the sub-telomeric regions (striped blue bars). The trend observed by Brauer et al. is thus presumably not due to slower growing cells spending more time in M-G1 phase as suggested by the authors, Instead, it is likely an artifact of the many Y’ helicase genes found in the sub-telomeric regions of budding yeast, which are so highly homologous that they can cross hybridize on microarrays and hence all appear to be periodically expressed with identical peak times.

After correcting for this the down-regulated genes show a weak preference for being expressed during M phase whereas the up-regulated genes tend to be expressed in late G1 and S phase. However, the peak-time distributions of up- and down-regulated do not differ significantly from that of all cell-cycle-regulated genes (Kolmogorov-Smirnov test).

In summary, my reanalysis suggests that there is no correlation between the transcriptional response to changes in growth rate and transcriptional cell-cycle regulation. It also reiterates the importance of manually inspecting the results from statistical analyses – they may be highly significant for all the wrong reasons.

Cite this post

When I launched this blog just a month ago, I wrote that it was an experiment and that I would later evaluate if it was worthwhile to continue. I am very happy to already now be able to declare the experiment a success. With over 2000 views during the first month and around 50 subscriptions to the RSS feed, Buried Treasure is certainly worthwhile writing – I hope you also find it worthwhile reading.

In the near future I plan to experiment with live blogging. The idea is to provide coverage of the conferences and meetings that I attend. To avoid polluting my blog with dozens of short posts consisting of just one or two sentences and a photo, I have setup an associated Picasa web album. This is where the main action will happen – I plan to post only highlights and summaries here on the blog. You can find the RSS feeds for the album and the blog in the sidebar.

“Oh, you work on systems biology? So do I!”

New buzzwords to describe scientific disciplines and technologies seem to pop up every year. For the fun of it, I have developed a small web resource, BuzzClouds, that provides a visual overview of the latest buzzwords in biomedicine.

Without destroying your weekend with mathematical formulas, here is how the BuzzCloud selection and visualization method works:

• A list of potential buzzwords is constructed by extracting all one- and two-word phrases ending on -ics, -ology, -omy, -phy, -chemistry, -medicine, or -sciences. These endings were select to get buzzwords that correspond to scientific disciplines and technologies.
• The potential buzzwords are ranked according to a score that takes into account their frequencies within the past year and within the preceding decade (for details see this review article). To get a high score, a buzzword must be both frequent and new. The top-50 buzzwords are included in the cloud.
• The size of each buzzword is proportional to the logarithm of its frequency during the past year. Common buzzwords are thus large where as rare buzzwords are small.
• The brightness of each buzzword shows the frequency of the buzzword within the past year relative to the preceding decade. New buzzwords are thus bright whereas the older ones are darker.
• Finally, each buzzword is assignd a tint that goes from yellow via white to cyan based on how often it occurs in scientific journals (yellow) as opposed to medical journals (cyan).

When run for the year 2007, the end result looks like this (BuzzClouds for other years are available from the web resource):

I think the method does a pretty decent job despite the occasional mistakes such as nice technology and timely topics. In terms of scientific buzzwords, quantitative proteomics is booming, systems biology still hot although it is getting a bit long in the tooth, and synthetic biology is rapidly gaining popularity. And nanotechnology seems to be popular within the medical domain, giving rise to buzzwords like nanomedicine and nanotherapeutics.

Maybe I should write a buzzword-compliant, interdisciplinary grant application that combines click chemistry and synthetic biology to develop novel nanotherapeutics.

Cite this post

I have previously blogged about the relationship between cell-cycle phenotypes and regulation in human as well as budding yeast. I was thus excited to see the new RNAi study on cell-cycle phenotypes by Rines and coworkers that was published in Genome Biology two days ago. The title of their paper is “Whole genome functional analysis identifies novel components required for mitotic spindle integrity in human cells”, and the abstract reads as follows:

Background

The mitotic spindle is a complex mechanical apparatus required for accurate segregation of sister chromosomes during mitosis. We designed a genetic screen using automated microscopy to discover factors essential for mitotic progression. Using a RNAi library of 49,164 double-stranded (ds)RNAs targeting 23,835 human genes, we performed a loss-of-function screen looking for siRNAs that arrest cells in metaphase.

Results

Here we report the identification of genes that when suppressed result in structural defects in the mitotic spindle leading to bent, twisted, monopolar or multipolar spindles and cause a cell cycle arrest. We further described a novel analysis methodology for large-scale RNAi datasets which relies upon supervised clustering of these genes based on gene ontology (GO), protein families, tissue expression and protein-protein interactions.

Conclusions

This approach was utilized to functionally classify the identified genes in discrete mitotic processes. We confirmed the identity for a subset of these genes and examined more closely their mechanical role in spindle architecture.

The screen identified a set of 226 genes that when suppressed lead to spindle-related cell-cycle phenotypes. Using the name-mapping files from STRING, I was able to map 175 of them to the set of genes used in my other cell-cycle analyses. The results presented below are all based on this set of 175 genes.

To my surprise, Rines and coworkers did not compare their results to the earlier phenotypic screen published by Mukherji et al. in PNAS. Since I had already mapped this dataset onto the same gene set, it was easy to make a comparison of the new phenotype data from Rines et al. and the eight phenotypic categories defined by Mukherji et al.:

The statistical significance of the overlap was assessed using Fisher’s exact test and the false discovery rate (FDR) was calculated using the Benjamini-Hochberg method. As can be seen, the agreement between the two studies is very poor. Nonetheless, it is reassuring that the largest overlap (>8%) is observed for category 8, since spindle defects should be expected to result in problems during cytokinesis.

I also looked into the transcriptional and post-translational regulation of the 175 genes. The cell-cycle microarray study by Whitfield and coworkers covered 124 of the genes, 15 of which are periodically expressed (P < 0.002; Fisher’s exact test). Plotting the distribution of peak times for these genes confirms the observation by Rines et al. that the genes tend to be expressed around the G2/M transition and during M phase:

As should be expected, the peak-time distribution for the genes identified by Rines et al. is in agreement with the corresponding distributions for categories 4, 5, and 8 from Mukherji and coworkers.

Comparison with a set of 985 phosphoproteins identified in low-throughput studies (obtained from Phospho.ELM) shows that the proteins products encoded by the 175 genes are preferentially phosphorylated (P < 0.001; Fisher’s exact test). This result is confirmed by comparisons with large mass-spectrometry studies (P < 0.03; Fisher’s exact test) and CDK substrates predicted by NetPhosK (P < 0.05; Fisher’s exact test).

Finally, I analyzed the protein products encoded by the 175 genes for degradation signals. 22 of them contain a strong D-box motif (P < 0.03; Fisher’s exact test) and 28 contain a KEN-box motif (P < 0.002; Fisher’s exact test). By contrast, the gene products identified by Rines et al. display no overrepresentation of PEST degradation signals. This makes sense since proteins with D-box and/or KEN-box motifs are polyubiquitinated by the anaphase-promoting complex (APC) during late M phase, which targets them for degradation by the proteasome.

In summary, Rines and coworkers has identified a set of genes that show weak but significant overlap with some of the phenotypic categories defined by Mukherji et al., with periodically expressed genes identified based on microarray data from Whitfield et al., with known and predicted phosphoproteins, and with predicted degradation signals. All of the results are consistent with the majority of the 175 genes functioning during G2/M and early M phase.

Cite this post

Together with collaborators in Søren Brunak’s group, I have earlier published a comparative study on eukaryotic cell-cycle regulation. In the supplement and earlier papers, we presented benchmarks that documented the sensitivity with which periodically expressed genes can be identified based on microarray expression data. We thereby showed that the poor evolutionary conservation of transcriptional cell-cycle regulation is not an artifact of individual gene lists being unreliable.

However, there is a more direct test that we did not think of at the time, namely to check if the changes in periodic transcription agree with the binding of cell-cycle transcription factors in each organism. The first step is to select two organisms (organism 1 and organism 2) and extract two sets of genes: 1) cycling genes from organism 1 with non-cycling orthologs in organism 2 and 2) non-cycling genes from organism 1 with cycling orthologs in organism 2. Next, Fisher’s exact test is used to determine if targets of cell-cycle transcription factors are overrepresented in the first set relative to the second. This procedure is equivalent to the test for coevolution between transcriptional and postranslational regulation (see Jensen et al. (2006) for details).

I used the procedure to perform all pairwise tests for Homo sapiens, Saccharomyces cerevisiae, Schizosaccharomyces pombe, and Arabidopsis thaliana. For each choice of organism 1, I used the same set of cell-cycle transcription-factor targets also used for the original benchmarks. The table below sumarizes the results of the statistical tests; the rows specify organism 1 and columns specify organism 2:

For most of the pairwise organism comparisons, the expected coevolution of transcription factor binding and cell-cycle-regulated transcription is supported by the statistical test. Benjamini-Hochberg correction for multiple testing was thus not performed as it would change the p-values only marginally (by a factor of 4/3 to be exact). Apart from the S. pombe vs. S. cerevisiae comparison, the weak correlations all involve A. thaliana for which only very limited microarray expression data is available.

This analysis shows that the differences in cell-cycle-regulated transcription (as measured by microarrays) are consistent with the available data on transcription-factor binding. This provides direct evidence that the poor conservation of cell-cycle regulation observed between eukaryotes is due to genuine, biological differences.

Cite this post

Attila Chordash over at “PIMM – Partial immortalization” discovered that Proteomics have now changed the abstract of the infamous paper by Warda and Han to be an apology to their readership:

While I am pleased to see this public apology from the publisher, the retraction is still only based on “a substantial overlap of the content of this article with previously published articles in other journals”. That is a euphemism for “the authors copied four entire pages of text from sources that were not cited”. However, I am concerned that this apology – like the press release from Proteomics – ignores the central question: how did the manuscript make it through peer review?

I was a bit surprised to see an apology being published via PubMed, but a quick search revealed that Proteomics is far from the only journal to apologize to their readers in this way. In fact, a systematic count of the abstracts mentioning the words “apologise(s)” or “apologize(s)” has increased exponentially over the past decade (note the logarithmic scale):

The number shown for 2008 is an extrapolation based on the first six weeks; if the apologies keep coming at the current rate, there will be 32 by the end of the year. The line shows an exponential fit of the data points from 1999 to 2007. The doubling time for the number of apologies is just 3 years whereas the number of papers doubles only every 22 years. If these trends continue, there will be more apologies than papers published from the year 2067 and onwards. I apologize for the extrapolation.

Cite this post

In 2006 the Schultz lab at the Scripps Research Institute published a paper in PNAS called “Genome-wide functional analysis of human cell-cycle regulators”. The abstract reads:

Human cells have evolved complex signaling networks to coordinate the cell cycle. A detailed understanding of the global regulation of this fundamental process requires comprehensive identification of the genes and pathways involved in the various stages of cell-cycle progression. To this end, we report a genome-wide analysis of the human cell cycle, cell size, and proliferation by targeting >95% of the protein-coding genes in the human genome using small interfering RNAs (siRNAs). Analysis of >2 million images, acquired by quantitative fluorescence microscopy, showed that depletion of 1,152 genes strongly affected cell-cycle progression. These genes clustered into eight distinct phenotypic categories based on phase of arrest, nuclear area, and nuclear morphology. Phase-specific networks were built by interrogating knowledge-based and physical interaction databases with identified genes. Genome-wide analysis of cell-cycle regulators revealed a number of kinase, phosphatase, and proteolytic proteins and also suggests that processes thought to regulate G1-S phase progression like receptor-mediated signaling, nutrient status, and translation also play important roles in the regulation of G2/M phase transition. Moreover, 15 genes that are integral to TNF/NF-κB signaling were found to regulate G2/M, a previously unanticipated role for this pathway. These analyses provide systems-level insight into both known and novel genes as well as pathways that regulate cell-cycle progression, a number of which may provide new therapeutic approaches for the treatment of cancer.

I recently wrote a commentary about how phenotypes in yeast agree remarkably well with the just-in-time assembly hypothesis for cell-cycle regulation of protein complexes. I thus decided to also compare the dataset on cell-cycle phenotypes for human genes with the cell-cycle microarray expression data published in 2002 by Whitfield and coworkers.

Using the mapping files from the STRING database, I was able to automatically map 741 of the 1152 genes with cell-cycle phenotypes to the set of 12,097 genes for which we have cell-cycle microarray expression data. Of the 741 genes, 55 are among the a of 600 periodically expressed genes identified in a reanalysis of the data from Whitfield and coworkers. This is just shy of 50% more than what would be expected by random chance (P < 0.001; Fisher’s exact test).

The authors divided the cell-cycle mutants into eight classes. Repeating the above analysis for each of these categories separately revealed that genes with phenotypes related to S-phase and cytokinesis were significantly overrepresented among the 600 periodically expressed genes (FDR < 0.05; Fisher’s exact test and Benjamini-Hochberg correction for multiple testing). The other categories did not yield statistically significant results.

To look at the temporal regulation of transcription in more detail, I plotted the distribution of peak times (the point in the cell cycle when a gene is maximally expressed) for the periodically expressed genes from each of the eight phenotypic categories:

For the periodically expressed genes that display a cell-cycle phenotype in the screen by Schultz and coworkers, the observed phenotypes agree with the time of peak expression. In particular, the genes with cytokinesis-related phenotypes are all expressed shortly before the time of cell division (cytokinesis). Most of the periodically expressed genes with phenotypes related to S phase are similarly expressed during S phase (roughly 50-70% into the cell cycle), genes with phenotypes related to the G2/M transition also tend to be expressed during the appropriate phase of the cell cycle.

In summary, these results support the view that cell-cycle-regulated genes are expressed shortly before their time of action, despite the fact that regulation also takes place at the protein level. It also confirms that many genes with cell-cycle function are not subject to transcriptional cell-cycle regulation.

Cite this post

There is good news regarding the Warda and Han scandal. After numerous researchers including myself emailed the Editor in Chief of Proteomics, Michael J. Dunn, the paper is now listed as retracted. I am pleased to see that the editorial team of Proteomics has acted swiftly against plagiarism.

Edit: The last author of the paper, Jin Han, has written a reply to PZ Myers. According to the email, he has himself contacted the editorial office and requested that the paper be retracted. I am still looking forward to hearing an official explanation from Proteomics of how this paper got accepted in the first place.

Edit: Michael J. Dunn has emailed me a copy of the approved press release from Proteomics that announces the retraction of the paper by Warda and Han. The only explanation offered is that the paper made it through peer review due to “human error” – or in other words “someone did something wrong”. I would have been truly worried if a paper like this had been accepted without human error being involved. I hope that Proteomics will provide the scientific community with more details when they have completed the internal investigation of the incident.

Cite this post

The law of diminishing returns is a well known concept in economics. Highly simplified, it states that as you invest more, the overall return on investment increases at a declining rate. I wondered if this principle applies to biomedical research.

I thus wrote a small script to parse the Medline database and count for each year 1) the number of new papers published, 2) the number of authors that published at least one paper, and 3) the total number of (co-)authorships. The plot below shows the number of new papers and the number of active authors for each year since 1970:

Few scientists – if any – will be surprised to see that the rate of publication and the number of active publishing scientists have increased exponentially. However, it is slightly disconcerting that the number active authors doubles every 17 years whereas the number of papers per year doubles only every 22 years.

To look deeper into this, I plotted as function of time the average number of coauthors per paper and the average number of papers coauthored by each active author:

These two measures also appear to increase exponentially. However, the number of coauthors per paper is increasing considerably faster than the number of papers coauthored by each author per year. The estimated doubling times are 33 years for the number of coauthors per paper and 63 years for the number of papers coauthored. This suggests that the productivity of biomedical scientists, measured in terms of publications, has decreased.

A more direct way to show this is to plot the ratio between the number of papers published each year and the number of authors on them (note that the y-axis does not start at zero):

The fact is that the number of papers produced per researcher per year has dropped by roughly one third since 1970. However, there could be many reasons for this:

• Have we simply become lazy?
• Has the bar been raised for what is considered the Least Publishable Unit?
• Are large collaborations less efficient than smaller projects?
• Do we spend more time on bureaucracy and less time on science?
• Or are we left with the hard questions because the easy ones have all been answered?

My guess is that the last three reasons all play important roles. What do you think?

Cite this post

This post might be considered off topic since it is about a paper that is unfortunately neither buried nor treasure. I would rather describe it as “organic fertilizer that has come into contact with a rotary air-circulation device”.

The paper that I will dissect is “Mitochondria, the missing link between body and soul: Proteomic prospective evidence” by Mohamad Warda and Jin Han. This review is at the time of writing published in electronic form by the journal Proteomics (ISI impact factor 5.735). It was aptly described as “A baffling failure of peer review” by PZ Myers on his blog Pharyngula, which led to a flash mob of researchers (including me) quickly identifying several flaws, any one of which should in my view be sufficient to cause the journal to retract the paper:

• Warda and Han twice suggest that mitochondria provide a link between body and soul, but they never provide any argument for this.
• They claim to present data that disproves the accepted endosymbiotic theory for the origin of mitochondria. But in reality they present no such evidence.
• They promise to replace this theory by “a more realistic alternative”. The alternative turns out to be “a mighty creator”, or in other words Intelligent Design.
• To support these truly remarkable claims, the authors misrepresent the results of cited references. Some of these references are even completely unrelated to the topic at hand.
• Entire sections or paragraphs of the paper are plagiarism of other researcher’s papers and from the website of another group. Not only is this material not presented as quotations, the sources are not even cited.
• Finally, numerous sentences have been copied verbatim from the cited sources. This may be partly excused by the authors borrowing better English, but I nonetheless consider it an unacceptable practice especially in reviews.

Note how I use block quotations below to show which parts are not my own words. That is what Warda and Han should have done to avoid their biggest problem: being accused of plagiarism. Let us start by looking at the first sentence of the abstract:

Mitochondria are the gatekeepers of the life and death of most cells that regulate signaling, metabolism, and energy production needed for cellular function.

This sentence is identical to the first sentence on the webpage of a competing group, namely the Mitochondrial Research & Innovation Group at University of Rochester Medical Center. The rest of the paragraph from the webpage can be found later in the paper by Warda and Han:

Recent scientific studies show that mitochondrial dysfunction is more commonplace than previously thought and that substantial mitochondrial involvement is present in many acute and chronic diseases. Mitochondrial dysfunction is now implicated in a range of human diseases, including aging, diabetes, atherosclerosis, heart failure, myocardial infarction, stroke and other ischemic-reperfusion injuries, neurodegenerative diseases including Alzhiemer’s and Parkinson’s diseases; cancer, HIV; sepsis and trauma with multiorgan dysfunction or failure. Some rare mitochondria diseases (e.g., MELAS, Kearns-Sayre) are associated with large deletions in the mitochondrial genome. More recently, the so-called OXPHOS diseases that reflect a limited capacity to produce the energy needed to respond to normal stress conditions.

However, most of the plagiarized material is not from webpages; it is from peer-reviewed papers of other researchers. For example, the two paragraph below appear to originate from the paper “Peroxisome Proliferator-Activated Receptor gamma Coactivator-1 (PGC-1) Regulatory Cascade in Cardiac Physiology and Disease” published by Brian N. Finck and Daniel P. Kelly in the journal Circulation:

Emerging evidence supports the notion that derangements in mitochondrial energy metabolism contribute to cardiac dysfunction [186]. For example, human mitochondrial DNA mutations resulting in global impairment in mitochondrial respiratory function cause hypertrophic or dilated cardiomyopathy and cardiac conduction defects [187, 188]. Mutations in nuclear genes encoding mitochondrial fatty acid oxidation enzymes may also manifest as cardiomyopathy [189, 190]. Interestingly, cardiomyopathies resulting from inborn errors in mitochondrial fatty acid oxidation enzymes are often provoked by physiological or pathophysiological conditions that increase dependence on fat oxidation for myocardial ATP production such as prolonged exercise or fasting associated with infectious illness [190,191].

A causal relationship between mitochondrial dysfunction and cardiomyopathy is also evidenced by several genetically engineered mouse models. Targeted deletion of the adenine nucleotide translocator 1, which transports mitochondrially generated ATP to the cytosol, leads to mitochondrial dysfunction and cardiomyopathy [192]. Mice with cardiac-specific deletion of the transcription factor of activated mitochondria, which controls transcription and replication of the mitochondrial genome, also exhibit marked impairments in mitochondrial metabolism, severe cardiomyopathy, and premature mortality [193]. Cardiomyopathy and/or conduction defects are also observed in several mouse models with targeted deletion of specific fatty acid oxidation enzymes [194, 195].

I discovered these plagiarized sections myself, but they only scratch the surface and pale in comparison to the amount of copied material identified by others. I should make clear that I do not blame the reviewers for not discovering this; their job is to check the scientific quality of the material presented, not to detect fraudulent or plagiarized material.

The editor and the reviewers are not off the hook, though. Interspersed between the sensible review material, much of which has been copied from elsewhere, there are a few sections that are “a mélange of truths, half-truths, quarter-truths, falsehoods, non sequiturs, and syntactically correct sentences that have no meaning whatsoever” (to use the words of Alan Sokal).

The first of these is the following sentence from the abstract:

These data are presented with other novel proteomics evidence to disprove the endosymbiotic hypothesis of mitochondrial evolution that is replaced in this work by a more realistic alternative.

Clearly the editors and the reviewers should have examined the evidence for such an exceptional claim. The “evidence” that supposedly disproves the serial endosymbiotic theory (SET) of mitochondrial evolution is presented in section 3.4, which after explaining the theory makes the following baffling statement:

The proof of SET was based on the parallel connection between plant mitochondrial and phage T4 genome replication [107, 108] …

This is simply not true. First, the scientific method can never prove a theory. Second, the similarity between the replication of mitochondrial and phage T4 genomes is completely irrelevant with respect to the endosymbiotic origin of mitochondria. Third, reference 108 is about chloroplasts and not mitochondria.

Next, the authors try to convince the reader that people are still debating the validity of the endosymbiotic theory:

Therefore, the debates concerning the mitochondrial endosymbiotic hypothesis recently terminated with many questions still left unanswered [111].

Reference 111 is a paper in Cellular Immunology by Gray and coworkers with the title “Modulation of CD8+ T cell avidity by increasing the turnover of viral antigen during infection”. It has nothing to do with the evolution of mitochondria.

The authors go on to recite the well known facts that the vast majority of mitochondrial proteins are encoded by nuclear genes and that most bacterial genes do not match within the mitochondrial DNA. They then give a long description of how tightly integrated mitochondria are with the rest of the cell and use the tired old argument of irreducible complexity to dismiss the endosymbiotic theory. Needless to say, the authors have what they consider to be a more realistic explanation:

Alternatively, instead of sinking in a swamp of endless debates about the evolution of mitochondria, it is better to come up with a unified assumption that all living cells undergo a certain degree of convergence or divergence to or from each other to meet their survival in specific habitats. Proteomics data greatly assist this realistic assumption that connects all kinds of life. More logically, the points that show proteomics overlapping between different forms of life are more likely to be interpreted as a reflection of a single common fingerprint initiated by a mighty creator than relying on a single cell that is, in a doubtful way, surprisingly originating all other kinds of life.

In other words: “God did it”. If I can read correctly, the authors here reject not only the endosymbiotic origin of mitochondria but also the common ancestry of eukaryotes. And as if this was not enough, the authors end the paper with the following bold conclusion that also explains the mysterious title of the paper:

We realize so far that mitochondria could be the link between the body and this preserved wisdom of the soul devoted to guaranteeing life.

I am speechless. As anyone who knows me can attest, that very rarely happens.

Edit: The paper has now been retracted, but there are still many open questions as to how it got accepted in the first place.

Cite this post

Beginning of this year Ben Lehner’s lab published a beautiful study in BMC Systems Biology with the title “A simple principle concerning the robustness of protein complex activity to changes in gene expression”. The abstract reads:

Background

The functions of a eukaryotic cell are largely performed by multi-subunit protein complexes that act as molecular machines or information processing modules in cellular networks. An important problem in systems biology is to understand how, in general, these molecular machines respond to perturbations.

Results

In yeast, genes that inhibit growth when their expression is reduced are strongly enriched amongst the subunits of multi-subunit protein complexes. This applies to both the core and peripheral subunits of protein complexes, and the subunits of each complex normally have the same loss-of-function phenotypes. In contrast, genes that inhibit growth when their expression is increased are not enriched amongst the core or peripheral subunits of protein complexes, and the behaviour of one subunit of a complex is not predictive for the other subunits with respect to over-expression phenotypes.

Conclusions

We propose the principle that the overall activity of a protein complex is in general robust to an increase, but not to a decrease in the expression of its subunits. This means that whereas phenotypes resulting from a decrease in gene expression can be predicted because they cluster on networks of protein complexes, over-expression phenotypes cannot be predicted in this way. We discuss the implications of these findings for understanding how cells are regulated, how they evolve, and how genetic perturbations connect to disease in humans.

It struck me that these observations can all be explained by the just-in-time assembly model for temporal regulation of protein complex assembly, which I developed together with members of Søren Brunak’s group. For a long explanation and discussion of the model see our paper “Evolution of Cell Cycle Control: Same Molecular Machines, Different Regulation”. For the short version see the figure below, which shows how cell-cycle regulation of just a single subunit is sufficient to control when during the cell cycle a complex is active (click to enlarge):

What will happen if you knock down the expression of one subunit of a complex? The maximal number of complete complexes that can be assembled will be reduced, irrespective of whether the subunit is dynamic or static. Whether this results in a given phenotype depends on the function of the complex. However, the effect should in principle be the same for different subunits of the same complex, which is exactly what Lehner and coworkers observed.

What if you instead overexpress one subunit of a complex? For a static subunit it should not really matter; the maximal number of complete complexes that can be assembled is unchanged. On the other hand, overexpression of a dynamic subunit may cause the complex to become constitutively active, which could have disastrous consequences for the cell. Overexpression of dynamic and static subunits of the same complex should thus give rise to different phenotypic effects. This would explain the observation by Lehner and coworkers that subunits of the same complex often have different overexpression phenotypes.

If this hypothesis is true, genes that lead to phenotypic effects when overexpressed should preferentially encode dynamic proteins, i.e. many of the genes should be periodically expressed. In fact, this correlation between overexpression phenotype and cell-cycle regulation was already described by the Hughes, Boone and Andrews labs who originally published the dataset on overexpression phenotypes (for details see their paper in Molecular Cell):

Genes expressed periodically during the cell cycle (de Lichtenberg et al., 2005) were more likely to show an overexpression phenotype (p = 0.017), and in particular, this tended to cause abnormal morphology [p < 10^-13] or cell cycle arrest [p < 10^-14](Table S3). When the analysis is limited to genes known to function in the mitotic cell cycle, we still find that overexpression of periodically expressed genes is more likely to cause cell cycle arrest (p = 0.008) or abnormal morphology (p = 0.006) than constitutively expressed cell cycle genes (Table S3), indicating that unscheduled expression of genes that are usually expressed periodically often leads to toxicity.

The results of the two papers thus point in the direction that the just-in-time assembly hypothesis can explain the qualitatively differences between knock-down and overexpression phenotypes.

Cite this post

A while back, I found a paper in PLoS Genetics from Jason Lieb’s lab entitled “Cell Cycle-Specified Fluctuation of Nucleosome Occupancy at Gene Promoters”. There is little point in me rephrasing their work, so here is the original abstract:

The packaging of DNA into nucleosomes influences the accessibility of underlying regulatory information. Nucleosome occupancy and positioning are best characterized in the budding yeast Saccharomyces cerevisiae, albeit in asynchronous cell populations or on individual promoters such as PHO5 and GAL1–10. Using FAIRE (formaldehydeassisted isolation of regulatory elements) and whole-genome microarrays, we examined changes in nucleosome occupancy throughout the mitotic cell cycle in synchronized populations of S. cerevisiae. Perhaps surprisingly, nucleosome occupancy did not exhibit large, global variation between cell cycle phases. However, nucleosome occupancy at the promoters of cell cycle–regulated genes was reduced specifically at the cell cycle phase in which that gene exhibited peak expression, with the notable exception of S-phase genes. We present data that establish FAIRE as a high-throughput method for assaying nucleosome occupancy. For the first time in any system, nucleosome occupancy was mapped genome-wide throughout the cell cycle. Fluctuation of nucleosome occupancy at promoters of most cell cycle–regulated genes provides independent evidence that periodic expression of these genes is controlled mainly at the level of transcription. The promoters of G2/M genes are distinguished from other cell cycle promoters by an unusually low baseline nucleosome occupancy throughout the cell cycle. This observation, coupled with the maintenance throughout the cell cycle of the stereotypic nucleosome occupancy states between coding and noncoding loci, suggests that the largest component of variation in nucleosome occupancy is ‘‘hard wired,’’ perhaps at the level of DNA sequence.

Although the authors clearly and very reasonably focus on nucleosome assembly at the promoter regions, their microarray experiments include also probes for the open reading frames (ORFs). I thus decided to take a look at whether nucleosome occupancy within the ORFs correlates with the transcriptional regulation of genes.

I first downloaded the complete dataset from the Gene Expression Omnibus (GEO) database and used a previously described algorithm to identify genes with periodic variation in nucleosome occupancy. The algorithm provides two p-values that tell if a profile varies significantly across time points (“regulation”) and if this variation correlates significantly with a cosine wave with the period of interest (“periodicity”). Finally, it combines the two p-values into a single score (“combined”).

The genes can be ranked according to any one of these three scores. To test which score makes most biological sense, I benchmarked each of the ranked list against a list of 113 genes that are known from small-scale experiments to be transcriptionally regulated during the budding yeast cell cycle (the B1 list from de Lichtenberg et al. (2005)). Below is a plot showing the fraction of the benchmark identified as function of the number of genes identified as having periodic nucleosome occupancy:

As should be the case, the curve for the combined score lies above the curves for “regulation” and “periodicity” alone, and all three scores enrich for know cell-cycle-regulated genes relative to random expectation (“random”). A clear break in the “combined” curve is observed at rank 300, after which there is essentially no enrichment for known cell-cycle-regulated genes. I thus decided to base my analysis on the top-300 genes according to the combined score.

Comparing this list to the top-600 periodically expressed genes from de Lichtenberg et al. (2005) revealed an overlap of just 45 genes. This is approximately 50% more than expected by random chance and the difference is statistically significant (P < 0.001; Fisher’s exact test).

The next logical thing to do was to check the timing of expression during the cell cycle for the genes with and without periodic nucleosome occupancy. The temporal expression profile of each periodically expressed gene is summarized in a single number, the peak time, which tells when in the cell cycle the gene is maximally expressed. The unit of the peak times are “percent of a cell cycle”; 0% corresponds to the time of cell division and 40% corresponds to S phase (in budding yeast). I plotted the peak-time distributions as histograms:

If the two distributions look similar to you, it is for a good reason: according to the Kolmogorov-Smirnov test there is no significant difference. Not a very exciting result, but perhaps also not too surprising.

If the periodic nucleosome occupancy of cell-cycle-regulated genes has a biological imporantance, then one should expect that time of peak expression of a gene corresponds to the time of minimal nucleosome occupancy. I thus made a scatter plot of the two for the set of 45 genes:

There seems to be no correlation. A possible explanation is that overlap of 45 genes is only 50% more than expected by random chance; in other words only one in three genes, that is 15 genes, can be expected to contribute to the signal (if there is any). Even if the correlation was perfect these genes, the overall correlation would be difficult to detect.

At this point I decided to drop the project. In summary, the benchmark results and the comparison with microarray expression data show that there is a statistically significant correlation between periodic nucleosome occupancy within an ORFs and periodic expression of the corresponding gene. However, the signal is quite weak and it is thus difficult to get much further.

You are most welcome to post good ideas as comments. Alternatively, I will be happy to provide you with all the files if you want to take over the project.

Cite this post

I’m a computational biologist who tends to work on way too many projects at any one time. Many of these result in observations that I find interesting – but not interesting enough to bother writing up a manuscript and sending it to a peer-reviewed journal. In other cases the results were simply negative and no conclusions could be drawn. My disk may thus contain “buried treasure”.

My primary goal with this blog is to make my never-to-be-published observations openly available. As I don’t plan on continuing these projects, anyone is welcome to pick up a project and continue where I left off. I also plan on reporting negative results on this blog so that others can hopefully avoid repeating analyses that would lead them nowhere.

Please note that this blog is an experiment. Over the next few months I will try to write up a number of posts on various projects. After a test period, I will make up my mind about whether to continue or not. This depends on how many people read and comment my posts vs. how long it takes me to write the posts. Like for all other experiments, time will tell if it becomes a success.