Public Rambling

Predicting kinase specificity from phosphorylation data

noreply@blogger.com (Pedro Beltrao) — Wed, 25 May 2011 17:37:00 +0000

Over the past few years, improvements in mass-spectrometry methods have resulted in a big increase in throughput for the identification of post-translational modifications (PTMs). It is even hard to keep up with all the phosphoproteomics papers and the accumulation of phosphorylation data. Most often, improvements in methods result in interesting challenges and opportunities. In this case, how can we make use of this explosion in PTM data ? I will try to explore a fairly straightforward idea, on how to use phosphorylation data to predict kinase substrate specificity. I'll describe here the general idea and just the first stab at it to show that I think it can work.

The inspiration for this is the work by Neduva and colleagues that have show that we can search for enriched motifs within proteins that interact with the domain of interest. For example, we can take a protein containing and SH3 domain, find all of it's interaction partners and you will likely see that they are enriched for proline rich motifs of the type PXXP (x = any amino-acid) that is the known binding preference for this domain. So the very obvious application to kinases would be to take the interaction partners of a kinase and find enriched peptide motifs. The advantage of looking at kinases, over any other type of peptide binding domains, is that we can focus specifically on phosphosites.

As a test case I picked the S.cerevisiae Cdc28p (Cdk1) that is known to phosphorylate the motif [ST]PXK. I used the STRING database to identify proteins that functionally interact with Cdc28 with a cut-off of 0.9 and retrieved all currently known phosphosites within these proteins. As a quick check I used Motif-X to search for enriched motifs. The first try was somewhat disappointing but after removing phosphosites that had less than 5 MS spectra and/or experiments supporting it I got back the this logo as the most enriched motif:

This was probably the easiest kinase to try since it is known that it typically phosphorylates its targets at multiple sites and it heavily studied. Still, I think there is a lot of room for exploration here. If anyone is interested in collaborating on this let me know. If your doing computational work I would be interested in some code/tools for motif enrichment. If your doing experimental work let me know about your favorite kinases/species.

In defense of 'Omics

noreply@blogger.com (Pedro Beltrao) — Thu, 28 Apr 2011 21:29:00 +0000

High-throughput studies tend to have a bad reputation. They are often derided as little more than fishing expeditions. Few have summarized these feelings as sharply as Sydney Brenner:

"So we now have a culture which is based on everything must be high-throughput.I like to call it low-input, high-throughput, no-output biology"

Having dealt with these type of data for so long, I am often in the strange position of having to defend the approaches. As I was in a real need to procrastinate, I decide to try to write some of these thoughts down.

Error rates
One of the biggest complaints directed at large-scale methods is that they have very high error rates. Usually these complaints come from scientists interested in studying system X or protein Y, that dig into these datasets only to find out that their protein of interest is missing. Are the error rates high ? While this might be true for some methods it is important to note that the error rates are almost always quantified and that those developing the methods keep pushing the rates down.

When thinking about 'small-scale' studies I could equally ask - why should I trust a single western blot image ? How many westerns were put in the garbage bin before you got that really nice one that is featured in the paper ? In fact, some methods for reducing the error become only feasible when operating in high-throughout. As an example, when conducting pull-down experiments to determine protein-protein interactions, unspecific binding becomes much easier to call. This has lead to the development of analysis tools that cannot be employed on single pull down experiments.

So, by quantifying the error rates and driving these down via experimental or analysis improvements, 'omics research is in fact, on the forefront of data quality. At the very least, you know what the error rate is and can use the information accordingly. Once the methods are improved to an extent that the errors are negligible or manageable they are quietly no longer consider "omics". The best example of this I think is genome sequencing. Even with the current issues with next-gen sequencing, few put 'traditional' genome sequencing in the same bag with the other 'omics tools, although they have quantifiable errors.

Standardization
Related to error quantification is standardization. To put is simply, large-scale data is typically deposited in databases and is available for re-use. What is the point of having really careful experiments if they will only be available for re-use, in any significant way, when a (potentially sloppy) curator digs the info out of papers ? This availability fuels research by others that are not set-up to perform the measurements. This is one of the reasons why bioinformatics thrives. The limitations become the ideas not the experimental observations/measurements. Anyone can sit down, think of a problem and with some luck the required measurements (or proxy of them) have been made by others for some unrelated purpose. This is why publications of large-scale studies are so highly cited, they are re-used over and over again.

Engineering mindset and costs
One other very common complaint about these methods is cost. It is common to feel that 'omics research is 'trendy', expensive and consumes too much of the science budgets. While the part about budget allocation might be true, the issue with costs is most certainly not. Large-scale methods are developed by people with an engineering mindset. The problems in this type of research are typically on how to make the methods work effectively, which includes making them cheaper, smaller, faster, etc. 'Omics research drives costs down.

Cataloging diversity
Besides these technical comments the highest barrier to deal with, when discussing these methods with others is a conceptual one. Is there such a thing as 'hypothesis free' research ? To address this point let me go off on a small tangent. I am currently reading a neuroscience book - Beyond Boundaries - by Miguel Nicolelis, a researcher at Duke University. I will leave a proper review for some later post but, at some point, Nicolelis talks about the work of Santiago Ramon y Cajal. Ramon y Cajal is usually referred to as the father of the neuron theory that postulates that the nervous systems is made up of fundamental discrete units (neurons). His drawings of neuronal circuits of different species are famous and easily recognizable. The amazing level of detail and effort that he put into these drawings really underscores his devotion for cataloging diversity. These observations inspired a revolution in neuroscience, much the same way Darwin's catalogs of diversity impacted biology. Should we not build catalogs of protein-interactions, gene-expression, post-translational modifications, etc ? I would argue that we must. Omics research drives errors and price down, creates catalogs of easily accessible and re-usable observations that fuels research. I actually think that it frees researchers. While a few specialize in method developments others are free to dream up biological problems to solve with the data gathering effort shortened to a digital query.

Miss-understandings
So why the negative connotations ? Part of it is simple backlash against the hype. As we know, most technologies tend to follow a hype cycle where early exaggerated excitement is usually followed by disappointment and backlash when they fail to deliver. A second important aspect is simply a lack of understanding of how to make use of the available data. This model of data generation separated from the problem solving and analysis only makes sense if researchers can query the repositories and integrate the data into their research. It is sad to note that this capacity is far from universal. While new generations are likely to bring with them a different mindset, those developing the large scale methods should also bear the responsibility of improving the re-usability of the data.

Structure based prediction of kinase interactions

noreply@blogger.com (Pedro Beltrao) — Thu, 03 Mar 2011 20:26:00 +0000

About a year ago Ben Turk's lab published a large scale experimental effort to determine the substrate recognition preferences of most yeast kinases (Mok et al. Sci. Signal. 2010). They used a peptide screening approach to analyze 61 of about 122 known S. cerevisiae kinases in order to derive, for each one, a position specific scoring matrix (PSSM) describing their substrate recognition preference. In the figure below I show an example for the Hog1 MAPK where it is clear that this kinase prefers to phosphorylate peptides that have proline next to the S/T that is going to be phosphorylated.

Figure 1 - Example of Hog1 substrate recognition preference derive from peptide screens. Each spot in the array contains a mixture of peptides that are randomized at all positions except at marked position (-5 to +4 relative to the phosphorylatable residue). Strong signal correlates with a preference for phosphorylating peptides containing that amino-acid at the fixed position.

As was previously known, most kinases don't appear to have very striking substrate binding preferences. Still, these matrices should allow for significant predictions of kinase-site interactions. These matrices should allow us also to benchmark previous efforts by Neil and other members of the Kobe lab on the structural based predictions of kinase substrate recognition. For this, I obtained the predicted substrate recognition matrices from the Predikin server and known kinase-site interactions from the PhosphoGrid database. I used this data to compare the predictive power of the experimentally determined kinase matrices (Mok et al.) with the predicted matrices from Predikin. This analysis was done about a year ago when the Mok et al. paper was published but I don't think Phosphogrid was significantly updated since then.

Phosphogrid had 422 kinase-site interactions for the 61 kinases analyzed in Mok et al. of which ~50% of these have in-vivo evidence for kinase recognition. As expected, the known kinase-site interactions have a stronger experimental matrix score than random kinase-site assignments (Fig 2).

Figure 2 - The set of kinase-site interactions used broken down according the kinases with higher representation. These sites were scored using the experimental matrices along with other randomly selected phosphosites and the scores of both populations are summarized in the boxplots.

A random set of kinase-phosphosite interactions of equal size was used to quantify the predictive power of the experimental and the Predikin matrices with a ROC curve (Fig 3).

Figure 3 - Area under the ROC curve values for kinase-site predictions using both types of matrices.

Overall, the accuracy of the predicted matrices from Predikin matched reasonably well with those derived from the peptide array experiments with only a small difference in AROC values. I broke down the predictions for individual kinases with at least 10 sites known. Benchmarking of such low numbers becomes very unreliable but besides the Cka1 kinase, the performance of the Predikin matrices matched reasonably well the experimental results.

I am assuming here that Predikin was not updated with any information from the Mok et al study to derive their predictions. If this is true it would mean that structural based prediction of kinase recognition preferences, as implemented in Predikin, is almost as accurate as preferences derived from peptide library approaches.

Why would you publish in Scientific Reports ?

noreply@blogger.com (Pedro Beltrao) — Fri, 07 Jan 2011 03:40:00 +0000

The Nature Publishing Group (NPG) is launching a fully open access journal called Scientific Reports. Like the recently launched Nature Communications, this journal is online only and the authors cover (or can choose to cover for Nat Comm) the cost of publishing the articles in an open access format. Where 'Scientific Reports' differs most is that the journal will not reject papers based on their perceived impact. From their FAQ:
"Scientific Reports publishes original articles on the basis that they are technically sound, and papers are peer reviewed on this criterion alone. The importance of an article is determined by its readership after publication."

If that sounds familiar it should. This idea of post-publication peer reviewing was introduced by PLoS ONE and Nature appears to be essentially copying the format from this successful PLoS journal. Even the reviewing practices are the same whereby the academic editors can choose to accept/reject based on their opinion or consult external peer reviews. In fact, if I was working at PLoS I would have walked into work today with a bottle of champagne and I would have celebrated. As they say, imitation is the sincerest form of flattery. NPG is increasing their portfolio of open access or open choice journals and hopefully they will start working on article level metrics. In all, this is a victory for the open-access movement and to science as a whole.

As I had mentioned in a previous post, PLoS has shown that one way to sustain the costs of open access journals with high rejection rates a publishers needs also to publish higher volume journals. Both BioMedCentral and more recently PLoS have also shown that high-volume open access publishing can be profitable so Nature is now trying to get the best of both worlds. Brand power from high-rejection rate journals with a subscription model and a nice added income with a higher-volume open access journals. If by some chance, founders force a complete move to immediate open access, NPG will have a leg to stand on.

So why would you publish in Scientific Reports ? Seriously, can someone tell me ? Since the journal will not filter on perceived impact, they wont be playing the impact factor game. They did not go as far as naming it Nature X so brand power will not be that high. It is similarly priced (until January 2012) as PLoS ONE and has less author feedback information (i.e. article metrics). I really don't see any compelling reason why I would choose to send a paper to Scientific Reports over PLoS ONE.

End of the year with chemogenomics

noreply@blogger.com (Pedro Beltrao) — Fri, 31 Dec 2010 17:14:00 +0000

Taken from jurvetson at:
www.flickr.com/photos/jurvetson/3156246099/

Around this time of the year it is customary to make an assessment of the year that is ending and to make a mental list of things we wish for in the year ahead. Here is my personal (but work related :) take on this tradition.

My academic year ended with the publication of two works related to chemogenomics. Chemogenomics or chemical genomics tries to study the genome-wide response to a compound. Usually, collections of knock-outs or over-expression of large number of genes are grown in the presence or absence of a small molecule to assess the fitness cost (or advantage) of that perturbation to the drug response. This is what was done in these two works.

In the first one, Laura Kapitzky (a former postdoc colleague in the lab) used a collection of KO strains both in S. cerevisiae and S. pombe to essay for the growth in the presence of different compounds. The objective was to study the evolution of the drug response in these distantly related fungi. In line with what was previously observed in the lab for genetic-interactions and kinase-substrate interactions we found that drug-gene functional interactions were poorly correlated across these two species. Perhaps one interesting highlight from this project was that we could combine data from both fungi to improve the prediction of the mode-of-action of the compounds.

The second project, in which I was only minimally involved in, was a similar chemogenomic screen but at a much larger scale. As the tittle implies "Phenotypic Landscape of a Bacterial Cell" (behind paywall), is a very comprehensive study of the response of the E.coli whole knock-out library against an array of compounds and conditions. Robert, Athanasios and other members of the Carol Gross lab did an amazing job of creating this resource and picking some of the first gems from it.

Something that I wanted to highlight here was not so much what was discovered but what I was left wanting. These sort of growth measurements tell us a lot about drug-gene relationships. We also have a growing knowledge of how genes genetically interact either by similar growth measurements in double-mutants or by predictions (as in STRING). These should allow us then to make prediction about how drugs interact. If two drugs can act in synergy to decrease the growth of a bug we should be able to rationalize that in terms of drug-gene and gene-gene interactions. I find this is a very interesting area of research. Naively these sort of data should allow us to predict drug combinations that target a specific species (i.e. pathogen) or diseased tissue but not the host or the healthy tissue. Here is a scientific wish for 2011, that these and other related datasets will give us a handle on this interesting problem.

As for the future, I am entering the final year of my current funding source (thank you HFSP) so my attention is turning into finding either some more funds or another job. I will continue working on the evolution of signalling systems, in particular trying to find the function of post-translational modifications (aka P1). Unfortunately the project failed as an open science initiative, something that I have mostly given up for now. I think the main reason it didn't work was because of lack of collaborators of similar (open) interests and non-overlapping skill sets as Greg and Neil were discussing in the Nodalpoint podcast a while ago.

See you all in 2011 !

The GABBA program

noreply@blogger.com (Pedro Beltrao) — Tue, 21 Dec 2010 22:36:00 +0000

I was recently in the annual meeting of my former PhD program, the GABBA program, a Graduate Program in Areas of Basic and Applied Biology in Portugal. I realized that I never blogged about the Portuguese PhD programs and I thought I would share with you their somewhat unusual concept.

Like in other PhD programs, GABBA students start by having courses during the first semester of the program. The semester is divided into week long courses in different subjects (think Cell-Cycle, Development, etc) with invited teachers. What is different from most other programs I know of is that students then get to use their scholarship to do their research projects anywhere in the world. GABBA students get payed to do their research in any lab that accepts them, no strings attached. No return clause, not even a requirement to inform the program of research progress. There is an annual meeting where students (and alumni) get to go to Portugal to present their work but no one is obliged to go. It is also a nice opportunity to exchange tips and in some cases even start collaborations.

The annual meeting is always organized around Christmas time so most people end up going. I kept going to the meetings after finishing my PhD mostly because I enjoy seeing the people but also because of the cool science. As you can imagine, everyone is scattered around the world in very nice labs doing research in all sort of different biomedical related subjects. This year there were a lot of talks about stem cells and an unusually high number of neurobiology related work. Some cool research of note for me were for example the work of Martina Bradic (Borowsky lab at NYU) about the convergent evolution of blind cave fish and the talk by Andre Sousa (Sestan Lab at Yale) on the transcriptional profiling of human brain regions during development (http://hbatlas.org/).

The GABBA program takes international students as well but they are typically asked to do their research in Portugal. The applications are usually around June so keep an eye out if you are interested in applying. Have a look at the admissions page for more information.

This holiday season, make them spit in a tube

noreply@blogger.com (Pedro Beltrao) — Wed, 24 Nov 2010 21:44:00 +0000

Black Friday is upon us and everyone here in the US is going consumer crazy. Along with the traditional discounts in the offline world, there are also tempting promotions in many online stores. One great example is the discount that 23andMe is offering until next Friday. If you have not heard about 23andMe, they are a direct-to-consumer genetics company that sell a SNP profiling service. You get to find out about your ancestry and genetic propensity for traits and some diseases. The analysis usually costs $499 (plus a one year $5 monthly mandatory subscription) but they are having a $400 dollar discount (use promo code UA3XJH). What better way to spend Christmas than having everyone spit into a little tube.

ICSB 2010 - From design principles to genome-wide and that persistent gap in-between

noreply@blogger.com (Pedro Beltrao) — Wed, 20 Oct 2010 00:43:00 +0000

I am back from the 11th International Conference on Systems Biology (ICSB 2010) that was held in the lovely cite of Edinburgh. A full week of talks dedicated to systems biology (however you might define it). Speaking of definitions, it was refreshing to note that no speaker spent much time trying to define the field this time around. Aside from the keynote lectures, there were constantly four parallel sessions (see program) so you were always guaranteed to miss out on something interesting. With the help of a couple of attendees we took some notes of the conference in a FF room. There is lots of notes to go through if you are interested but here a few of my favorite highlights.

Keynotes
From all the keynotes the ones I enjoyed most were from Luis Serrano, Mike Tyers and Aldons Lusis. Unfortunately, Sydney Brenner was scheduled but canceled last minute.

Luis Serrano talked about his lab's work on characterizing a mycoplasma species (notes here) using a multi-omics approach. They are trying to use available technologies to build parts lists and models of all aspects of this species. They have learned a lot by taking such a systematic approach on a bug with so few genes but there is no plan that I could see on how to follow up on all this work. I can see that many computational biology labs will use this data but it would be a missed opportunity if more labs don't continue to go beyond the omics limitations.

The lecture by Tyers (notes here) was also very much about omics. He talked about their recent effort (Breitkreutz et al Science 2010) to search for kinase-protein interactions in yeast and how hard it is, in general, to study signalling pathways in this way (promiscuous interactions, complex systems etc). From kinases he moved to drug-gene interactions and chemogenomics. In particularly he briefly mention some unpublished work on evolution and prediction of drug-synergy. This is topic that I am really interested in as an applied side of evolutionary biology (more on that hopefully soon).

Another keynote I enjoyed was from Aldons Lusis (see notes). His presentation centered on a strategy for association studies in mice (Bennett et al. Genome Research 2010). This sort of work is out of my comfort zone but I really like all the examples he gave of using this strategy to find loci associated with clinical traits or protein/gene expression levels. Maybe I should be trying to read Nature ~~Association Studies~~ Genetics more often.

Parallel sessions
I went to the sessions of "Functional Genomics", "Cell signalling dynamics","Parameterizing proteomics" and "Biological noise and cell decision making". In the Functional Genomics session, Lars Steinmetz talked about genome wide analysis of antisense non coding transcription and David Amberg's talk covered the use of genetic interactions to study actin mutants (an extension from Haarer et al. G&D 2007).
I really liked many of presentations from the Cell Signaling Dynamics session, including talks by Walter Kolch, Timothy Elston and Nils Bluthgen. It was interesting to note that many people presenting were following a similar approach of first enumerating different models that could achieve the function they were studying and then finding the most plausible by elimination.
From the proteomics session the highlight for me was the really cool work presented by Christian von Mering. They have essentially compiled a lot of mass-spec data and used corrected spectral counts to estimate protein abundance for many different species. The data can be found at http://pax-db.org and some reported results are published (Schrimpf et al., PLoS Biology 2009 and Weiss et al., Proteomics 2010). Overall the message appears to be that protein abundance is more conserved across species than gene expression.
Finally, from the session on noise and cell decision I particularly liked the talk of Roy Kishony, on his lab's work with antibiotic response, and James Locke's analysis of sigma b promoters (Elowitz lab).

Bridging the gap
Besides all the cool science I come back from this meeting with the feeling that we still have this huge gap between the -omics work and the detailed 'design principles' pathway analysis. There is even such a tension between people working in these two camps that it becomes almost a joke. Maybe this is why it is so hard to define systems biology, each "type" of researcher sees it differently. Some would say that it is not systems biology if it is not genome wide, while others will claim that we don't learn anything with omics (just a parts list). In this meeting there were great examples of both camps using established methods to attack new systems but there is still no clear attempt to bridge the gap. How do we go from genome-wide to quantitative mechanistic understanding ? Maybe next year in Heidelberg / Mannheim (ICSB 2011) we will see both camps, at least, acknowledging each other.

Book Review: The Visual Miscellaneum

noreply@blogger.com (Pedro Beltrao) — Tue, 05 Oct 2010 05:03:00 +0000

After watching David McCandless's TED talk "The beauty of data visualization" I was sufficiently curious to go ahead and buy his most recent book "The Visual Miscellaneum". As expected, it is a very easy to "read" book containing interesting and beautifully presented trivia. The main idea behind the talk (and I assume the book) is to present information in a visual way to make data more tangible. McCandless complains that the information that is given in the news is hard to grasp without context and shows, through his infographics, how it can be improved. He keeps a website where some of the visualizations are freely available and if you enjoy them and the talk then the book is worth a look too.

What is interesting about these visualizations is that they occupy a space between art and data presentation. When I was going through the book I was considering if they could serve as a source of inspiration for research data. Is there room for more artistic visualizations in scientific articles ? Should we try to make data beautiful or does the primary objective of conveying the result totally overrides any artistic intention ? In a recent referee report a reviewer asked us to change some of the figures because he/she thought that there were redundant elements in them. The figure was not even about data but about the workflow of the project. This is not a complaint, just an example that probably illustrates well what is the current culture in academia. We are trained to be skeptical and constantly looking for tweaks or additions that obfuscate the results (i.e. weird scales, exploded pie charts). Maybe in our effort to be accurate we forget how important it is to make an image intuitive and pleasant.

Nodalpoint is back, as a podcast

noreply@blogger.com (Pedro Beltrao) — Wed, 22 Sep 2010 02:34:00 +0000

If you read this blog than there is a very high chance that you know about Nodalpoint. It was one of the first (community) blogs related to science and where many bioinformatic bloggers, myself included, started out. Over the years, the site lost usage as people started their own independent blogs and Greg Tyrelle, the creator of Nodalpoint, eventually archived it.

The main website is back, in a way. Greg decided to start up a podcast series to discuss issues around bioinformatics and I guess whatever else he might be interested in. Go check it out. The first episode is a conversation with Neil Saunders, one of Nodalpoint's early users (blog, friendfeed, twitter) .

Among many other things, they talk about the lack of traction that open science has among scientists. I agree with some of the points that were raised regarding the small size niche of each specific research problem. It is not the full answer but it probably plays a role. There are so few people that have the skills and interest to tackle the same problem that creating a online community around any given scientific question becomes hard. Still, if we have not come together to openly share results and methods we have at least witness the creation of many online communities that are working very well to discuss all sort of different scientific issues (ex. Friendfeed-Life Scientists, Biostart, Nature Network, etc).

Systems Biology versus "real" biology

noreply@blogger.com (Pedro Beltrao) — Fri, 17 Sep 2010 21:03:00 +0000

Scientific American has an article about this years' Lindau meeting of Nobel Laureates. It features an interesting conversations between Tim Hunt, Roland Pache (at the time PhD student) and undergraduate Sophia Hsing-Jung Li.
Here is the video of the conversation:
The discussion centered around systems biology and Hunt was not shy about expressing his skepticism. Since I happen to see great value in both the Omics and the design principles sort of work that characterize systems biology my frustration grew quickly. The whole video can be neatly summarize by Hunt's advice that people working in systems biology should "spend plenty of time talking to real biologists".

Real biologists ? ... I felt like writing a long rant about the findings that were made possible by the sort of work that he his so skeptical about but then I thought about xkcd and relaxed a bit:

Do we still need pre-publication peer-review ?

noreply@blogger.com (Pedro Beltrao) — Tue, 20 Jul 2010 22:15:00 +0000

A bit over a month ago Glyn Moody wrote a blog post arguing that abundance of scientific publishing outlets removes the need for our current system of pre-publication peer-review. The post sparked an interesting discussion here on FriendFeed.

Glyn Moody tells us that we have now:

"yet another case of a system that was originally founded to cope with scarcity - in this case of outlets for academic papers. Peer review was worth the cost of people's time because opportunities to publish were rare and valuable and needed husbanding carefully"

Since we have an endless capacity to publish information online Moody argues that there is no longer a need to pre-select before publication. We can leave that all behind us and do a post-publication peer-review that is distributed by all of the readers using all sorts of article level metrics that PLoS has been promoting.

More recently Duncan wrote another blog post that has some information that I think is important for this discussion. He was trying to estimate how many articles have ever been published. In the process he noted an interesting number - the number of articles that are currently published per minute. Pubmed keeps a table with the number of articles that they have information on per year. I don't think the last couple of years are well annotated and the first decades are that reliable so I just plotted here the totals between 1966 and 2007.

It is not surprising to see that the number of articles published per year is increasing, it probably matches well our expectations. I personally feel like I never have enough time to keep up with the literature. We are currently over the 700.000 papers per year. A search on pubmed for articles published in 2009 returns 848.856 papers. Something like 1.6 papers per minute !

So, although we have no scarcity of publishing outlets we have a huge scarcity of attention. It is very literally impossible to keep up with the current literature without some sophisticated filtering system. With all of the imperfections of our current System (TM) of editorial control, subjective peer review, subjective impact evaluations, impact factors and so on, we must agree that we need a lot of help filtering through these many articles.

I have read some people arguing that we should be capable ourselves of reading papers and realizing if they are interesting/innovative or not. That is fine for the very narrow range of topics that are close to our area of interest. I have pubmed queries for my topics of interest and I do filter through these myself without relying (too much) on the journal it was published on, etc. The problem is everything else that is not within this extremely narrow range of topics or the many papers that escape my queries. I want to be made aware of important new methods and new discoveries outside my narrow focus.

Moody and many others argue that we can do the filtering after publication by the aggregated actions of all of the readers. I totally agree, it should be possible to do the filtering after publication. It should be possible but it is not in place yet. So, if we want to do away with the System .. build a better system along side it. Show that it works. I would pay for tools that would recommend me papers to read. In my mind, this is where publishers of today should be making their money, in tools that connect the readers to what they want to read, not on content that should be free to read and re-use by anyone (open access).

Review - The Shallows by Nicholas Carr

noreply@blogger.com (Pedro Beltrao) — Thu, 15 Jul 2010 22:27:00 +0000

On a never ending flight from Lisbon back to San Francisco I finished reading the latest book from Nicholas Carr: "The Shallows - What the Internet is doing to our brains". The book is a very extended version of an article Carr wrote a few years ago enteitled "Is Google Making us stupid" that can be read online. If you like that article you will probably find the book interesting as well.

In the book (and article) Carr tries to convince the reader that the internet is reducing our capacity to read deeply. He acknowledges that there is no turning back to a world without the internet and he does not offer any solutions, just the warning. He explains how the internet, as many other communication revolutions (printing press, radio, etc), changes how we perceive the world. In a very material way, it changes our brain as we interact with the web and learn to use it. He argues that the web promotes skimming the surface of every web page and that the constant distractions (email, social networks) are addictive. This addiction can even be explained by an ancient species need to constantly be on the look out for changes in our environment. So, by promoting this natural and addictive shallow intake of information, the internet is pushing aside the hard and deep type of reading that has been one of mankind's greatest achievements.

After reading all of this I should be scared. I easily spend more than ten hours a day on these interwebs and my job as a researcher depends crucially on my capacity to read deeply other scientific works, reason about them, come up with hypothesis, experiments etc. So, why I am still writing this blog post instead of sitting in some corner reading some very large book ? Probably because I do not share Nicholas Carr's pessimist view. I actually agreed with a lot more things that I was expecting to before reading the book. I certainly believe that, like any other tool, the internet changes our brains as we used it. I agree also that reading online promotes this skimming behavior that the book describes. I observe the same from my own experience. What I find hard to believe is that the internet will result in the utter destruction of mankind as we know it (* unless saved by The Doctor).

It is just a personal experience but, despite my addiction to the internet, I haven't stopped reading "deeply". Not only is it a job requirement, I enjoy it. One of my favorite ways to spend saturday mornings is to get something to read and have long breakfast outside. At work I skim through articles and feeds to find what I need and when I do I print to read deeply. That is why I have piles of articles on my desk. This just to say that I found a way around my personal difficulty with deep reading on the computer screen. In other words, if it is required, we will find a way to do it. The internet habits that might be less conducive to deep thought are not worse than many any other addiction of our society and we have learned to cope with those.

I cannot imagine going back to a time when I would need to go to a library and painfully look for every single scientific article I wanted. Not to mention the impossibility of easily re-using other people's data and code. So even if a small but significant number of people can't find a way to cope with the lure of the snippets the advantages still overwhelmingly outnumber the disadvantages.

This topic and book have been covered extensively online. It is almost even evidence in itself that Carr is wrong that such a wealth of interesting and diverse opinions have shown up on the very technological platform that Carr is criticizing in the book (granted that some of these are also newspapers :). Examples:

Mind Over Mass Media (by Steven Pinker)
Carr's reply

Interview with Nick Carr and New York Time's blogger Nick Bilton

and for a different take on the topic here is an interview with Clay Shirky

Genetic interactions in powers of ten

noreply@blogger.com (Pedro Beltrao) — Thu, 27 May 2010 19:39:00 +0000

During my PhD at EMBL I attended a talk by Peer Bork where he said that computational biologists have the luxury of being able to work at any level of biological organization (atoms, cellular interactions, organism, ecosystems, etc) . At the time his lab was starting to work with metagenomics and his talks would cover the whole range of topics from protein domains to ecosystems. This idea of studying biology across this different scales reminded me of a very inspiring short movie entitled "Powers of Ten" (Wikipedia entry). This 1977 short movie was commissioned by IBM and it was written and directed by Ray Eames and Charles Eames. It takes the viewer on a journey in space from the very small atomic resolution to the outer reaches of the universe in incremental steps of powers of ten. Its only about 10 min long and if you haven't already seen here is below the embed version (while it lasts):

With all the different applications of genetic interaction screening going on here in the Krogan lab we though it would be interesting to write an essay that would, in the same spirit of this short movie, take the reader on a journey across different scales of biology. The essay was just made available online and I hope you enjoy the ride :).

We hope it serves as a tutorial for people interested in using genetic interaction data. There is more and more of this sort of data being deposited in databases and only a small fraction is being used to its fullest potential. We tried to show several examples of concrete findings that were first hinted by genetic interaction data.

Additionally we were trying to make the point that developments in high-throughout methods are decreasing the limitations of what can be observed in biological systems using the same methodologies. This is interesting because it challenge us to build models that can explain biological systems across different layers of biological organization. How does a change in DNA propagate across these layers ? Can it change the meaning of a codon, impact on a protein's stability/interactions, affect the action potentials in a neuronal cell and how species interact ? As we increase our capacity to monitor biological systems we should not only be able to tackle specific layers (i.e. understand protein folding) but we will eventually be concerned with coupling this different models to each other.

Kaggle - a home for data mining challenges

noreply@blogger.com (Pedro Beltrao) — Fri, 30 Apr 2010 02:46:00 +0000

I got a promotional email today from a new project called Kaggle. Somewhat related to Innocentive, this project aims to connect challenging problems with people that have the right set of skills to solve them. Kaggle is more specifically aiming to host prediction challenges and should appeal more to the data mining communities. For example, the site is currently hosting a challenge about HIV progression where problem solvers are giving a training dataset and asked to predict improvement in a patient's viral load.

I sent a few questions to Anthony Goldbloom (who works for Kaggle) to get a better idea of what the site is about:

Could you just tell me a bit about the company ?

The project was inspired by an internship I did as a journalist in London in 2008, when I wrote about the use of data by organizations. I am an econometrician by training and I was excited to see the principles we use to forecast economic growth, inflation etc, being applied by organizations. I returned to Australia and resolved to get involved in the broader analytics community. That's how Kaggle was born.

It looks like a young startup, is this right?

The project is only two weeks old and we've been thrilled with the response - we've attracted over 6,000 unique visitors.

We launched the Eurovision contest to get things going. In the last few days we released the HIV Progression Prediction competition. This was my introduction to bioinformatics, which seems like a fascinating area - we're hoping to attract more such competitions. Perhaps your readers have ideas or data?

Does the name mean anything ?

The name doesn't mean anything. I got tired of coming up with great names and finding they were taken (and that the owner would only sell for $xx,xxx). As a young project, our funds could be better spent elsewhere, so I built a program that iterated over different combinations of letters and printed a list of available and phonetic domain names. (I put this program on the web for others in a similar situation.)

How do you hope to be different from what Innocentive is doing ?

The project is solely focused on data competitions. This enables us to offer services - e.g. to help our clients frame their problems, anonymize their data, etc.

The platform is also easily extensible, so we can modify it to suit the specific needs of different data competitions.

We will host a rating system/league table, so that statisticians can use strong performances to market themselves. The rating system also allows us to host forecasting competitions, since the competition host will know who has a track record of forecasting well (and therefore who to pay attention to).

In the medium term, we plan to also offer a tender system, so that consultants can bid for work from organizations and researchers all over the world. From the organization's perspective, the rating system means they know what they're paying for. From the consultant's perspective, they don't have to waste time touting for work and they get access to interesting clients and datasets.

Science isn’t fair

noreply@blogger.com (Pedro Beltrao) — Tue, 27 Apr 2010 19:17:00 +0000

<rant>

Life isn’t fair, science is part of life therefore science isn’t fair. This would be a very short way to say what I am thinking but this is a rant so I will stretch it out a bit more.

We learn early on that in our line of work there is almost no correlation between the amount of work we do and the results we get. You need luck and I am not turning mystical on you here. I mean the low likelihood kind of luck. Even if you do everything right, being successful in science depends mostly on factors that are outside your control. A somewhat random pool of people end up being in the right place and the right time to go on with their academic work. Almost like a game of musical chairs, those with enough passion and perseverance to sustain the blows of lady luck get to play in the final rounds. Granted that I have been at this only for a few years but I have seen my share of hard working people getting scooped or hitting the wall with impossible projects. Try to explain scooping to non-scientists to see how ridiculous that sounds. I have also seen people (myself included) getting authorships for things I would not consider worthy of such.

So … science isn’t fair. This was exactly the sort of observations that made me start thinking about open science a few years ago. We could help to even out the playing field if we all are a bit more open about what we are working on. Too many financial and personal resources are eaten away to the duplication of research agendas.

</rant>

Nature Communications serves its first papers

noreply@blogger.com (Pedro Beltrao) — Tue, 13 Apr 2010 18:46:00 +0000

The new Nature brand journal (Nature Communications) has published its first set of papers this week. It is an interesting development in scientific publishing for many reasons. This is the first Nature brand journal that is online only and offers an (expensive) $5000 open access choice. Also, they are positioning this journal specifically as lower tier journal than previous Nature journals. According to the scope section:

"papers published in Nature Communications will be of high quality, without necessarily having the scientific reach of papers published in Nature and the Nature research journals."

So why is Nature dipping its toes in higher volume open access versus its typical market of highly selective closed access papers ? A bit of context might be required and some of the discussions from 2008 about the PLoS business model are worth revisiting. A few years ago, Declan Butler, a reporter from Nature, wrote an overly negative news piece about PLoS ONE which generated a huge online discussion (see Bora's link fest). Timo Hannay's reaction to this discussion was a much more balanced point of view from Nature's side of things. Essentially, Timo Hanny was pointing out that PLoS had failed to make a profit with their more selective journals and that it was showing that a lower tier of less selective journals are required to subsidize the higher tiers. Timo also said that PLoS was creating barriers to market entry for other OA publishers because they were using philanthropic grants to sustain their business.

So with this in mind, Nature Communications could be seen as bet hedging. Open access might be here to stay due to mandates from funding agency. If that is the case, the example from PLoS shows us that the only way to sustain highly selective journals is to publish also lower tier, less selective journals. This way the publishing house can also directly pass papers down its chain of journals and even possibly pass around the referee reports to expedite publishing.

If most publishers try to cover the whole range of journal selectivity how may publishers will there be a market for ?

While PLoS and Nature and expanding down this perceived pyramid of journal selectivity, BMC has been trying to expand up. This week, BMC Biology and Journal of Biology announced that these two journals are fusing to be the new flagship journal of BMC. I wish the best to the re-birth of BMC Biology but expanding up the ladder of "perceived impact" is much harder than expanding down.

Through this all we have still not managed to do away with this idea of journal prestige or impact. PLoS ONE promised us they would provide us with ways to filter and sort papers on their individual value but we are still not there yet. Ironically these "editorial" services might end up coming from third party programs like Mendeley, CiteUlike or Papers.

The stream

noreply@blogger.com (Pedro Beltrao) — Sun, 21 Feb 2010 08:29:00 +0000

http://www.flickr.com/photos/hamed/
CC BY 2.0

Google unveiled recently Yet Another Try at social networking in the form of Google Buzz. It is a social network borrowing heavily from Friendfeed, a website build by ex-googlers. If you are not familiar with Friendfeed here is a post that goes through some of its features.

One interesting thing about all this proliferation of social networks and feed aggregators is seeing their evolution over time. Over the past couple of years some of their features became somewhat standard. You could say that this is just because some websites keep stealing ideas from others but it also says which features seam to be useful and which implementations are intuitive to theirs users.

One idea that is central and common to all of these social websites is the concept of the stream. A list of updates from your contacts in the network typically ordered by time that you can interact with either by commenting or more simply by stating that you find that interesting. These actions are in turn propagated to your own contacts and so on.

It is impressive to see how this simple idea became so widespread in so little time. Facebook estimates that it has over 400 million active users. If Facebook was a country it would the 3rd most populous after China and India. We had plenty of ways to interact with friends and colleagues online before these social networks arrived (Email and instant messaging among others) so why did they become so popular ? The first few iterations of the stream reminded me a lot of those mass emails and chain emails from a few years back. It is also somewhat similar to how people were using their status in instant messaging tools to broadcast news about themselves. These two examples show that when given the tools people enjoy telling their contacts what their up to.
Status in instant messaging have no history and broadcasting jokes by email is very impolite as most people use email for work. So broadcasting to your social network in an non-intrusive way fills a need that previous tools could not solve well before.

Its clear that the stream is here to stay but where is it heading to ?

The stream localized
GPS enabled phones let us track our position and share it with the world. I am personally not comfortable with this but plenty of people are using tools like Foursquare and now Google Buzz to share their coordinates. In Foursquare users can play games where they "check-in" to places to unlock tips and badges. For business owners this could be used to give rewards for loyalty to their costumers.
It is easy to imagine how interesting it would be to get tips on what to eat when "checking in" to a restaurant or finding out that your friend is just around the corner in a cafe you like. Still, you don't have to be too paranoid to start thinking about the implications of telling the world where you are. "Please Rob Me" is the name of a website that, as the name implies, was created exactly to raise awareness to these privacy concerns.

Most likely these tools will iterate through changes in their privacy settings. For example, Google Latitude lets you share your location only to a select group of people or applications as well as letting you set the level of detail shared (ex. exact position versus area/city). Given the many business opportunities around location based advertisement companies will certainly try to make location sharing a standard property of the stream. The advertisement system in the movie Minority Report comes to mind.

Social Searching
After releasing Google Buzz the company also announced that they had acquired the company Aardvark. If you use sites like twitter or many of the other social networks you probably tried to broadcast a question. If you are not sure who exactly knows the answer there is no harm and casting a wide (and non-intrusive) net to try to find an answer. The term "lazy web" describes this sort of question broadcasting. In twitter there are even simple services organized around these "lazyweb" questions (see Lazytweet as exanple).

Aardvark tries to take this concept a bit further by targeting your questions to people that are more likely to known the answer instead of simply broadcasting to all your network. When you sign up to the service you tell it what subjects you might be able to answer and how often you mind getting some questions. In return you can ask Aardvark any question you want and it will try to route it to an "expert". This sort of social searches are a useful complement to current search engines. Your not supposed to ask questions that are easy to find with Google and it will take longer to get a reply but you can ask more subjective questions and hopefully get very knowledgeable answers.

I have tried asking questions in different social networks and a few times in Aardvark. Predictably the quantity and quality of the replies depends mostly on how specific the question is. Very broad and subjective questions get many useful replies while questions on very specialized topics will probably go unanswered.
The success of such an approach depends on many different factors but it looks like an interesting direction for search.

What do you think ?
In what other ways will we be using the stream ?

Review - You are not a gadget

noreply@blogger.com (Pedro Beltrao) — Fri, 05 Feb 2010 04:07:00 +0000

I just finished reading "You are not a gadget" by Jaron Lanier. The book is very much in the same tone as an article he recently wrote the Edge called "DIGITAL MAOISM:
The Hazards of the New Online Collectivism". Very few other books made me want to say "No!" out loud so many times while reading it. I enjoy reading opinions that run contrary to my own because I think it is important to challenge our ideas. This is why I like reading Rough Type. This book, however, was extremely confusing too me. It reads mostly as a collection of essays and often deviates from the path. I still think it was an interesting book to read because of the importance of the topic.

If you read the essay linked above you will get the general feeling conveyed in the book. As Lanier writes in the end of the first chapter:
"So, in this book, I have spun a long tale of belief in the opposites of computationalism, the noosphere, the Singularity, web 2.0, the long tail, and all the rest. I hope the volume of my contrarianism will foster an alternative mental environment, where the exciting opportunity to start creating a new digital humanism can begin".

I think these sentences summarize well what he set out to do in this book. To counter the rising open culture / web 2.0 movement and create some "alternative mental environment" for the future of the web culture. Some things he talks about I fully subscribe. If you believe that the singularity is near and that we are about to merge with the machines in the next couple of years you are about as bonkers as the rapture people. The wisdom of the crowds can do a great job at annotating images but it will not cure cancer. Also, the rise of the open culture (free content, mash-ups, etc) is hurting content producers and we can't just say that they are the dinosaurs and let them figure it out while we pirate their goods. Journalism is fundamental to democracy and we need to figure a way to make it work.

What I dislike about the book is the overly negative tone. How many people really believe that "wisdom of the crowds" can solve the worlds problems ? How many people have even heard of the term ? I would risk saying that Lanier spends too much time around silicon valley geeks. Sure, there is an open culture on the web but I pay more for content today that I ever did before (The Economist, Nature Reviews Genetics, Netflix, iTunes, Amazon on Demand, Pandora One, etc). The web 2.0 mash-up craze peaked when the Times nominated "You" as the person of the year (twitter is not content ;). Also, I like youtube clips like anyone else, some of them can be just amazing (ex. Kutiman's mash-ups) but I still want to pay to see Avatar again in glorious 3D IMAX.

One idea that he mentions often is that of the technological lock-in. As media formats might get locked in with use by the majority Lanier argues that concepts and ideas can be equally locked-in. An example he gives is the concept of files on the computers. That we are no longer free to experiment with the way information is stored in a computer system because this has been locked in.

What I guess Laniear was trying to say with this warning about technological lock-ins is that we run the risk of getting trapped in a set of ideas of the web that decrease the value of humanity and the content we produce and give too much value to the cloud of computers that underly the net. Even if I was to agree that current web culture tends to devalue content and humanity I don't think these lock-ins can be that powerful. We see net culture changing everyday before us and we have so far gained much more than we lost.

In summary, I would say that the problems he talks about are important but the book is overly pessimist about our current web culture.

Predicting and explaining drug-drug interactions

noreply@blogger.com (Pedro Beltrao) — Fri, 05 Feb 2010 02:40:00 +0000

I am generally interested in chemogenomic studies and drug interaction studies as a complement to what we work on in the Krogan lab (genetic interactions). Much like in genetic interaction screening, where the fitness of double mutant strains is compared with that of the individual single mutants, chemogenomics tries to identify drug-gene interactions while drug-drug interaction screening attempts to find cases where the combined effect of two compounds on fitness is different from the expected from the combination of the single independent effects.

I read two recent papers that I found interesting regarding drug-drug interactions. One was by Bollenbach and colleagues from the Kishony lab (published in Cell) and the other was by Jansen and colleagues (published in MSB). In the first, the authors present an explanation for a previously observed drug-drug interaction. It had been previously shown that the combination of DNA and protein synthesis inhibitors results in lower reduction of fitness than expected by a neutral combination model (termed antagonist interaction). The authors show in this paper that, in the presence of DNA synthesis inhibitors, ribosomal genes are not optimally expressed. This imbalance between ribosomal production and cell growth is detrimental to the cell and can be, at least in part, corrected by protein synthesis inhibitors, explaining why these can suppress the effects of the DNA synthesis inhibitors.

Although it is a relatively simple idea (once described), I think it shows how complex these drug-drug interactions can be and to some extent also how these can provide information about a cell.

In the second paper I mentioned, Jansen and colleagues try to develop an approach to predict drug-drug interactions based on chemogenomic data. There are many obvious reasons why this would be very useful and I find this line of research extremely interesting. What I was surprised with was the simplicity of the approach and the disappointing benchmarks.

The end-result from a chemogenomic screen is a vector of drug-gene interaction scores that tell us how the combination of a drug with each mutant (normally KO strains) affect growth when compared to neutral expectation from the combined effect of the individual perturbations. It had been previously shown that drugs that have a similar drug-gene vectors tend to have similar mechanisms of action (Parsons et al. 2006 Cell). What Jansen and colleagues now claim is that the similarity of drug-gene vectors are predictive not only of similar mode of action but also of drug-drug interactions. Specifically, they try to show that drugs with similar profiles are more likely to be synergistic, such that the combined effect of both drugs is expected to be more detrimental to the cell that the expected neutral combination.

Although the authors show experimental validation of their predictions with an accuracy of 56% they also benchmark their predictions using drug pairs previously known to be synergistic. This benchmark is somewhat disappointing since they only see a significant enrichment of these true-positive pairs for a narrow range of cut-offs and with 2 out of 3 ways of calculating drug-profile similarity. I wish the authors had comment on this difference between the relatively poor performance based on benchmark and the very high accuracy observed in their experimental tests. They also show that these predicted synergistic pairs are well conserved from S. cerevisiae to C. albicans which is contradictory to a previous Nature Biotech paper that I mentioned in previous post.

Are drug-synergies this easy to predict and so well conserved across species? I am personally not convinced based on the data from this paper alone so I am holding off for further validation by other groups or additional larger datasets/benchmarks.

Recently read - Jan 2010

noreply@blogger.com (Pedro Beltrao) — Fri, 22 Jan 2010 22:13:00 +0000

Blueprint for antimicrobial hit discovery targeting metabolic networks
Y. Shen, J. Liu, G. Estiu, B. Isin, Y-Y. Ahn, D-S. Lee, A-L. Barabási, V. Kapatral, O. Wiest, and Z. N. Oltvai

The authors use a flux balance analysis to identify reactions that essential for S. aureus growth. The enzymes required for the identified pathways were selected for in silico drug screening using both known structures and homology models. Inhibitors identified computationally where tend tested experimentally. I particularly liked the breath of different methods used in this study (FBA, homology modelling, ligand docking and experimental verification). It shows the usefulness of the increase in knowledge across these different areas (networks and structures).

Quantitative Phosphoproteomics Reveals Widespread Full Phosphorylation Site Occupancy During Mitosis
Jesper V. Olsen, Michiel Vermeulen, Anna Santamaria, Chanchal Kumar, Martin L. Miller, Lars J. Jensen, Florian Gnad, Jürgen Cox, Thomas S. Jensen, Erich A. Nigg, Søren Brunak, and Matthias Mann

In this study HeLa cells were synchronized in different stages of the cell cycle and their proteins and phosphorylation sites were quantified relative to asynchronous cells using a SILAC mass-spec approach. Changes in protein abundance and phosphorylation were combined with transcriptional changes and these were used to identify previously known and potentially novel complexes and kinase-substrate interactions important for cell-cycle progression. In addition, I thought it was pretty cool that the authors found a way to directly quantify the phosphorylation site occupancy from the mass-spec results. I was only slightly disappointed that the authors did not attempt to do a cross-species analysis given the available data from Liam Holt et al. on Cdk1 phosphorylation in S. cerevisiae. (Bonus- spot the blogger in the author list)

The Genetic Landscape of a Cell
Michael Costanzo et al.

This paper reports the large scale effort to quantify genetic interactions for approximately 1700 times 3800 pairs of genes in S. cerevisiae. As it is typically the case for these sort of "resource" papers describing a large dataset there is no way that a paper can do full justice to this work. They have mostly tried to show different ways to use this information: 1) predict gene function, 2) functional interactions between complexes and functional groups and 3) prediction of drug targets. Hopefully cell biology labs will pick up on this information to search for their genes of interested and bioinformatics groups will continue to find ways to make these resources easier to navigate (see STRING for a good example of this).

The joys of print

noreply@blogger.com (Pedro Beltrao) — Thu, 14 Jan 2010 22:14:00 +0000

For the past two months I have been enjoying my first ever print subscription to a scientific journal. The good folks over at Nature Reviews Genetics offered my a small discount that nudged me to it (thank you!). I thought that if I ever tried a print subscription it would be for sure a review journal. I can't say I regret the decision. Having the print issue to read on my commute makes me read articles that I would not normally print out and the front section (research highlights) is a good way to catch up to science news.

Maybe this explosion of e-readers will make it easier to emulate the browsing experience of a bound print copy. As Nicholas Carr (and others) have pointed out, changes in internet technology shape the way we think and use information. In a recent post, Carr gives his answer to this year's Edge annual question "How is the Internet changing the way you think?".
He writes:

"My own reading and thinking habits have shifted dramatically since I first logged onto the Web fifteen or so years ago. I now do the bulk of my reading and researching online. And my brain has changed as a result. Even as I’ve become more adept at navigating the rapids of the Net, I have experienced a steady decay in my ability to sustain my attention."

I think that Carr is, as usual, exaggerating in his pessimist view of web culture. In my work I scan online and print to read. The difference from that workflow to browsing a bound print copy is only the way I filter what I read. Still, it is worth a thought. Maybe we should be working on technologies that will help us work by forcing us to focus our attention better. For now, a print out will do :).

In science, data without purpose is sometimes required

noreply@blogger.com (Pedro Beltrao) — Mon, 11 Jan 2010 03:00:00 +0000

The title is probably flamebait but it might get you to read my little rant about data production in science. Its something I have been meaning to write about for a while but Deepak's post provided the extra incentive.

I think Deepak's post was a reminder that science is nothing without hypothesis and I certainly agree with that. To put this into context maybe it worth pointing again to the wired article about "The End of Theory" where Chris Anderson painfully tries to make the point that with the deluge of data that we are seeing we don't need models or hypothesis we just need to crunch the data to look for correlations.

I strongly disagree with this viewpoint. What would we learn about reality this way ? At most we would see correlations and could have some predictive power about future events but we would not know the mechanisms and thats the interesting part.

So why is data without purpose sometimes justified ? What I mean by this is that the capacity to produce data and its analysis does not have to be centralized in the same place. My perspective (bioinformatician) is from someone that has benefited a lot from the data deluge in biology and the fact that data is made (mostly) available to others. It has allowed many studies that reuse pre-existing results to answer new questions.

I also work in lab the develops genetic interactions screening methods and end having some discussions about this topic. Many people dislike this sort of research, finding creative names like "fishing expedition" to describe it. The truth is that there are many types of data that we need to collect (genomes, gene expression, protein-protein interactions, etc) that we know that will be useful to understand how cells work. We just need more accurate and cheaper methods to get them and there is no other way but to have the focus of the research be the data production itself.

Stitching different web tools to organize a project

noreply@blogger.com (Pedro Beltrao) — Sun, 03 Jan 2010 18:12:00 +0000

A little over a year ago I mentioned a project I was working on about prediction and evolution of E3 ligase targets (aka P1). As I said back then, I am free to risk as much as I want in sharing ongoing results and Nir London just asked me how the project is going via the comments of that blog post so I decided to give a bit of an update.

Essentially, the project quickly deviated from course since I realized that predicting E3 specificity and experimentally determining ubiquitylation sites in fungal species (without having to resort to strain manipulation) were not going to be an easy tasks.
So, since the goal was to use these data to study the co-evolution of phosphorylation switches (phosphorylation regulating ubiquitylation) it makes little sense to restrain the analysis specifically to one form of post-translational modification (PTM). After a failed attempt to purify ubiquitylated substrates the goal has been to come up with ways to predict the functional consequences of phosphorylation. We will still need to take ubiquitylation into account but that will be a part of the whole picture.

With this goal in mind we have been collecting for multiple species data on phosphorylation as well as other forms of PTMs from databases and the literature and we have been trying to come up with ways to predict the function of these phosphorylation events. These predictions can be broken down mostly intro tree types:
- phosphorylation regulating domain activity
- phosphorylation regulating domain-domain interactions (globular domain interfaces)
- phosphorylation regulating linear motif interactions (phosphorylation switches in disordered regions)

We have set up a notebook where we will be putting some of the results and ways to access the datasets. Any new experimental data and results from the analysis will be posted with a significant delay both to give us some protection against scooping and also to try to guarantee that we don't push out things that are obviously wrong. This brings us to a disclaimer... all data and analysis in that notebook is to be considered preliminary and not peer reviewed, it probably contains mistakes and can change quickly.

I am currently colaborating with Raik Gruenberg on this project and we are open to collaborators that bring new skills to the project. We are particularly interested in experimentalist working in cell biology and cell signalling that could be interested in testing some of the predictions we are getting out of this study.

I won't talk much (yet) about the results we have so far but instead mention some of the tools we are using or planning to use:
- The notebook of the project hosted in openwetware
- The datasets/files are shared via Dropbox
- If need arises code will be shared via Google Code (currently empty)
- Literature will be shared via a Zotero group library
- The papers and other items can be discussed in a Friendfeed group

This will be all for now. I think we are getting interesting results from this analysis on the evolution of the functional consequences of phosphorylation events but we will update the notebook when we are a bit more confident that we ruled out most of the potential artifacts. I think the hardest part about exposing ongoing projects is having to explain to potential collaborators that we intend to do so. This still scares people away.

I'll end with a pretty picture. This is an image of an homology model for the Tup1 -Hhf1 interaction. Highlighted are two residues that are predicted by the model to be in the interface and are phosphorylated in two different fungal species. This exemplifies how the functional consequence of a phosphorylation event can be conserved although the individual phosphorylation sites (apparently) are not.

Name that lab ...

noreply@blogger.com (Pedro Beltrao) — Thu, 17 Dec 2009 17:24:00 +0000

In the last editorial in Nature, the need for an author ID is introduced with the simple notion that each one of us has specific sets of skills:

In his classic book Management Teams, UK psychologist Meredith Belbin used extensive empirical evidence to argue that effective teams require members who can cover nine key roles. These roles range from the creative 'plants' who generate novel ideas, to the disciplined 'implementers' who turn plans into action and the big-picture 'coordinators' who keep everyone working together.

From this perspective the author ID is a tool that might help us get appropriate credit for skill sets that are currently undervalued. This sort of argument reminds me of a discussion I had several times in the past about the management structure of academic labs. Why is it that we have one single leader in each lab that has to handle all sorts of different management tasks ? Is it ego ? That we all need to have our own lab, named accordingly with our name ?

It does not take long to notice that all supervisors have their strengths and weaknesses and we talk about this openly. Some are better at grant writing, some have good people skills and keep the lab well balanced, a few (rare ones :) still know what they are talking about when they help you troubleshoot your method/protocol. If it was possible to have the same person doing all these things companies would not have come up with their more complicated management structures.

So why is it that we name our labs after ourselves and do a poor management job instead of having multiple PIs handling different aspects of the lab that is named after what it actually studies ?