Notes from the Biomass

The three major DNA sequence repositories (Genbank, EMBL, and DDBJ) decided to include for a location identifier in 2005. You can now access EMBL via a map of the world.

Not that I think that it is highly useful as of now but it gives a birds eye view of things to come.

The market for most specialized science software is rather small, consequently the software is usually expensive. Many scientists even expect scientific software to be transparent (or downright free the younger they are) but how would one sustain a company in a niche market?
Dan Gezelter from the OpenScience project is facing the dilemma and summarizes the points that may have been made before but have never been solved.
Giving the software away for free and try to live off consulting is an approach that has been tried by the people involved in MySQL or WordPress with Automattic.com. However, the small market size will make it a very risky business - after all, grad students are always cheaper than any outside consultant, even if they would work for food and lodging.

The post on the interaction paper that I found peculiar due its stress on the "first analysis" received several comments by email and verbally. Several of them were rather disappointed with the paper in general and stressed that the body of interactions used for inferences was very small and that little novelty was described in general. However, I find nothing "wrong" with the paper and if the editors find it interesting, why shouldn't they publish it?

A similar paper by Zhong and Sternberg in Science this Friday might receive the same criticism - much of it has been seen previously in other organisms, there is little novelty in the methodology. The authors integrate transcription data and interaction data from several organisms and provide experimental evidence for their prediction. The approach is not different from strategies employed by in the String database (Disclaimer: done by my former lab) or work from Lee et al. that also appeared in Science and several others. One might want to comment on the lack of references but given the severe restrictions by the journals it is problematic to select the right ones.
Further in defense of it: The work by Zhong and Sternberg might not seem to be novel in the eyes of the pundits but it is probably of considerable interest for the C. elegans community.

Other memes that often go around with papers from your core expertise in prestigious journals: "This work should not be in Nature Genetics" and "I wonder how it got into Science". Maybe I am just getting older but I relax and think that if these works are accepted, my work has a better chance of getting published too. At the end of the day, good work is recognized in your field even if it is not published in Cell. There are just less people jealous people commenting - is that a bad thing?

Or as a collaborator recently put it: You can get away with publishing good work in mediocre journals but not publishing mediocre work. Focus on your work, not the impact factor accrued by other scientists.

[Note to self: Re-read every Friday until you really believe it]

The analysis of the current data on protein-protein interactions described in human by TBK Gandhi et al in the March issue of Nature Genetics delivers a fair overview - solid but yielding little surprises. The authors compare human, yeast, fly and worm networks and verify a few transfers between species experimentally.

What struck me about the manuscript is the abstract which starts with:
We present the first analysis of the human proteome with regard to interactions between proteins.
Did the work by Lehner and Fraser not analyze human interactions - it is even cited? Many other publications analyzed aspects of the human interactome on many levels, e.g. work from text mining. There are works describing first glimpses into organisms that are not readily accessible for high-throughput methods. For instance, a recent computational analysis of Plasmodium falciparum does not claim to analyze the "interactome" but limits itself to prediction of new functions using protein-protein interactions.


Then again, every step to obtain a complete picture of biological data sets is oversold traditionally. The "complete human genome" was published about five times. There were only few experimental approaches to the problem of delivering a one-dimensional sequence. Given the complexity and the plethora of methods, the "complete human interactome" will be published a hundred times.

The stunning of images of simple bacteria growing in stunning patterns crossed my browser several times last week. Explore the astounding Sociofiction blog, one of the sources and other references from Pruned.

[via Kottke.org]

Gregory Petsko's editorial in Genome Biology compares the current practice of stuffing all data into tables and ill-formated text files as supplementary material and only giving an introduction and a discussion in the main paper with fighting a land war in Asia. I don't particularly like the analogy and the tone of the editorial in general but largely agree.
Currently, we attach complex data to a document style that is not much different from way science was communicated in 1950 despite advance in IT technology that make data sharing possible. However, we need to come up with better formats, not longer papers that include more data.

[Via Keats' telescope]