A physical view of networks
February 25, 2006
A simple physical model for scaling in protein–protein interaction networks
Authors: E. J. Deeds, O. Ashenberg, and E. L. Shaknovich
Journal: PNAS
Year: 2006
Volume: 103
Issue: 2
Pages: 311-316
ISSN: 0027-8424
This article caught my attention for a couple of reasons (1) It proposes a physical model for protein-protein interaction networks, and (2) It comes from Eugene Shaknovich, not someone I traditionally associate with protein-protein interactions, but more with folding and drug design.
“Scale-free” toplogies have been shown to exist in a number of networks, ranging from chemical reactions in a cell to the WWW. In general many biological systems seem to exhibit scale-free topologies. Deeds, et al. take a look at the network of protein-protein interactions (PPIs), specifically the results of two Y2H studies (Uetz et al., Nature, 2000 and Ito et al., PNAS, 2001) that suggest that the interactome of protein-protein interations of S. cerevisiae constitute scale-free networks.
The first thing that they observe is that there is little correlation between the two interactomes, which is not entirely suriprising given the noisy nature of Y2H experiments and the known false positive rates. This paper uses the hypothesis that Y2H experiments result in interactions that are dominated by non-specific interactions. They then proceed to propose a physical model (certainly something that interests me greatly) to explain how two uncorrelated networks might have such similar topologies. The core of the model proposed in this paper is that the surfaces get exposed randomly between experiments, exposing different sets of hydrophonic residues, which is sufficient to explain the lack of correlation.
The significance of this model is that PPIs as identified by Y2H are not necessarily driven by evolutionary dynamics, but rather by non-specific interactions (a factor supported by a correlation between the hydrophobicity of a protein and the number of interacting partners). It should be noted that this paper does not suggest that there is no biologically relevant and evolutionary information in the reported interactomes.
The approach in this paper is interesting from a protein modeling point of view, since it uses extrapolated homology modeling results as the basis for looking at surface hydrophobic residues. I find it interesting that the body of literature that tries to explain protein-protein interaction networks physically is relatively small. Perhaps I am biased, but the first thing that jumped out at me upon learning that the interactions between the same proteins in uncorrelated networks could not be random (due to the scale-free nature) was the question “what physical interactions are driving the interaction networks?”. In this case the authors propse that a simple hydrophobicity and de-solvation model (routinely used in protein-protein docking programs) + an element of noise can explain the differences in the results.
Is the model the explanation? I am not sure. Certainly it makes sense that protein-protein interactions are driven by hydrophobic interactions. It also would make logical sense that evolutionary pressure would drive proteins to interact via a specific set of residues, likely to be driven by desolvation and optimization of binding energies. The real picture is probably a little more complicated and has to balance a number of criteria. The model that the authors propose seems just a little too simple. It might just be an indictment of Y2H. If two experiments could identify highly correlated interactomes, I would not be surprised if they were driven by physical interactions, maybe evolutionarily conserved sets of surface patches that drive PPIs.
Further Reading:
Review at the bioinformatics blog
Technorati Tags: Systems Biology, Protein Interaction Networks, Science, Review
Someone actually wrote this?
February 17, 2006
Peter Cohen wrote this article for the Washington Post and I am still recovering from the shock of finding this on memeorandum. Not everyone likes mathematics or algebra, but just like history, geography and language, science and mathematics are absolutely essential for a well rounded education. As it is the US education system undervalues the level of mathematics and science education, and Peter is complaining about a year of algebra and geometry. Regardless of the utility (which is also unquestionable), being taught algebra and geometry results in a sharper mind, improved analytical skills, and plays an important part in developing the overall development of a child. Does being poor at mathematics mean that one is an idiot? Not at all, but it is as much a part of the overall education of a child as any other subject.
Not having been educated in the US, I can’t comment to the core reasons with any sense of certainty. However, If I had to guess, the reason is likely to be a combination of society undervaluing scientific and mathematical knowledge (from what I see sometimes ridiculing it) and flaws in the way mathematics and science are taught. For a country that has produced some of the finest scientific minds, and has a wonderful college system, the diminishing quality of science education at the school level is somewhat disconcerting
Further Reading:
The algebra haters club
Is algebra worthless?
Technorati Tags: Algebra, Mathematics, Society, Education
postgenomic: the memeorandum of the life science world
February 15, 2006
Stew of Flags and Lollipops fame has started a life science meme tracker called postgenomic. Still in beta, this is a wonderful idea and I recommend that anyone with the remotest interest in contemporary bioinformatics/genomics-based research bookmark the site.
Technorati Tags: Meme, Life Science, Bioinformatics, Genomics, Blogs, Information, postgenomic
Gold nanoparticles and bacteriophage - Sensors and cell-targeting
February 12, 2006
By now everyone probably knows how interested I am in using nanoscale assemblies for targeting cancer and other diseases. This particular paper is particularly fascinating as it uses a biological assembly as a model for a nanoengineered system. There are many applications here and number of lessons to be learned.
This is also a fascinating computational problem as it would require a true multi-scale modeling approach, incorporating molecular modeling, mesoscale modeling, continuum-scale modeling, informatics and an excellent visualization framework.
On a side note .. publicly available papers are such a wonderful thing.
Technorati Tags: Nanotechnology, Nanobiology, Nanomedicine, Self Assembly, Cancer, Healthcare
Evidence-based management .. a letter to the editor
February 11, 2006
Here is the text of a letter that I sent to the editor of Harvard Business Review.
Of all the articles I have read in HBR over the past couple of years, the article on Evidence-based management by Jeffrey Pfeffer and Robert Sutton was the first with enough issues to merit a letter to the editor. The first issue stems from the obviousness of much of the discussion, and the second from the poor choice of some of the examples. I am not a surgeon, but as a scientist I would argue that evidence-based approaches are the very fundamental basis of how someone trained in the sciences operates. The skills and knowledge that we learn in graduate school are form the backbone of our decisions over the course of our career. However as new evidence comes forward, this new knowledge has a significant impact on our decision making. I would argue that any scientist who does not keep pace with the literature and analyze all the evidence as it becomes available, is likely to be a poor one.
In addition many of the examples chosen are not the best to illustrate the success of evidence-based approaches. Yahoo is an example of that. It is no secret that many consider Yahoo’s front page to be a liability as it is cluttered and overloaded with information. The results from moving the search box to the center of the page would not qualify as an experiment, but rather an acknowledgement of the most successful search engine (in terms of use and deriving ad revenues), Google.
Even managers who make decisions using their experience are indirectly using evidence. In this scenario, experience is just another machine learning algorithm, wherein, a manager is able to examine all the evidence, and then use past experience (successes, failures, other examples) to extrapolate a decision using any new data presented. If there is a fundamental bias, then I doubt more analytical techniques would help in making the correct decision.
Technorati Tags: Management, HBR, Decision Making
