Symposium announcement: The Most Important Open Problems in Bioinformatics
Saturday March 25, 2006 CalIT2 auditorium, UCSD, La Jolla, CA.

If you are going to be in the area around this time I would suggest checking out this symposium coordinated by the UCSD Bioinformatics Graduate Program [full disclosure: this author is a member of this program]. The symposium is to be a survey of the big problems and current state-of-the-art from very prominent researchers around the country. Speakers include Eric Jakobsson (UIUC), Ruben Abagyan (Scripps), Kevin Karplus (UCSC), Sean Eddy (Wash U), and Adam Arkin (Berkeley).

http://abbey.ucsd.edu/symposium2006/

Yes, BTBR will be featuring reports from all the speakers. Hope to see you there!

A simple physical model for scaling in protein-protein interaction networks. by Deeds, Ashenberg, Shakhnovich in Jan 2006 PNAS.

I've been sitting on this review for a few days because I wanted to really think about what the authors are saying in this paper and figure out the best way to convey it. What this paper posits is that the touted scale free property of protein-protein interaction networks may be misinformed. Scale free network theory says that the structure of the network is evolutionary favorable and allows for modularity, hierarchy, and redundancy on the molecular level. What Deeds et al. are saying is that this scale free topology can be achieved using a completely non evolutionary method.

As many in the field well know, protein-protein interaction networks are full of false positives, and there is little correlation between published studies. The authors shed light on the reason for this by using a hydrophobic model for protein interaction. Pretty much if you just take the proteins with the most hydrophobic exposed surface area and give them a higher probability to interact with each other (non-specifically and more importantly in a non-biological context) you can reconstruct the scale free nature of published yeast protein interaction networks.

So what does this mean? Well I think that to some extent this paper is getting at one of the true underlying sources of false positives in interaction networks. In fact, this model may lead to a good way to prune out non-specific interactions. What this paper does not say, and I agree with them, is that there is no biological information in these studies. It's true that alone the data may be noisy, but compiled with other sources of information including expression, function, and conservation, the data remain extremely useful.




tt: Biophysics, Networks, Scale Free

Information Theory & Applications Inaugural Workshop at UCSD

Panel discussion: Robert Calderbank, Dave Forney, Michael Luby, Pavel Pevzner, Martin Vetterli, Andrew Viterbi, and Jacob Ziv.



Unfortunately, most of this discussion by premier information theorists was over my head. Robert Calderbank, the moderator, started off by asking the rest of the panel what the greatest triumphs and failures of the information theory field have been. Some were unsure that the field has advanced since Shannon. There was much discussion of CS and ECE topics, including the educational systems. One of them said that everybody should study physics as an undergrad. Physics gives you problem solving skills that every interdisciplinary team of scientists needs. Bioinformatics only came up when they discussed the future of information theory. Jacob Ziv said that biologists pretending to be bioinformaticians do not realize that all the technological tools they want have already been developed by the CS and ECE types, but they just don't know how to use them.

Pavel Pevzner expressed qualms about systems biology. He said that he still does not really understand what it is. He continued to say that he thought for a long time this was due to his own intellectual shortcomings, but then some other famous guy published a paper or review titled, "What is systems biology?" and then he felt relieved that he wasn't the only one. He also said that mathematicians frequently struggle with the motivation of their research, because they are working on problems that are interesting to them but may have no practical value to anyone else. So then Jacob Ziv said you just have to work on whatever you get a kick out of and hope that it will lead to something useful.

Andrew Viterbi said that any fame and accomplishments he has are only due to being in the right place at the right time.

What I took away from this conference was that to do real, applied bioinformatics, it is not enough to be a biologist writing shell scripts or a computer scientist trying to pipet or (even worse) working with data scrounged off somebody in the wet lab. You have to have a real command of both fields. When you generate your own data in the wet lab, you know how reliable they are and where the noise comes from, and you can design a more effective algorithm accordingly.

Also, before one of the talks I was shocked to see Andrew Viterbi sitting in my row, so I asked to take a picture with him. !!!!!!!

The metabolic world of Escherichia coli is not small, by M Arita (2004) PNAS. This brief communication illustrates that analyses of biological network topologies must always be cognizant of how the networks are constructed. By extracting atomic fluxes between metabolites from a genome-scale model of metabolism in E. coli, Arita concludes that the average distance between any two metabolites is much greater than suggested by earlier topological analyses which did not consider atomic fates. Ultimately, though we may try to deduce meaning from these increasingly thorough topological studies on biological networks, Arita's paper introduces something of possibly greater utility: a mass-centric network reconstruction method with homogeneous node types and physically meaningful edges. A reader is forced to wonder whether this method will generate as much insight as the better known flux-centric networks used to predict physiological bounds on metabolic, signaling, and even regulatory networks.

* written by guest reviewer