In case you weren’t aware, biology is really complicated.

It’s complicated not only because understanding specific proteins and pathways is difficult, but because understanding how a biological system of many proteins and pathways functions is even more difficult. In recent years, however, computer technology has made it possible to convert vast databases of biological information into an understanding not only of how individual genes and pathways work, but of how those individual genes and pathways work together.

The Diseasome project is one such project (and one of many -ome/omics words that you’ll encounter in the field of biology) which has converted human gene-disease relationship data from NCBI’s Online Mendelian Inheritance in Man (OMIM) into an annotated network exploring the genetic relationships between all the diseases covered in OMIM. The picture above is a small piece of the full poster (warning the file size is 20 MB) available at the site. An interactive version of the map is also available at the web page (and includes many links to Wikipedia entries explaining each of the genes and diseases) and is a fascinating browse-through for anyone who’s even remotely interested in how new network analysis techniques may be used in understanding human disease.

The Barbabasi Lab at Northeastern published an interesting paper in PNAS about how the data was compiled and analyzed for insights into how diseases and disease genes are connected and how that differs from how “normal” genes are connected. Although the work is subject to the standard Garbage-in-Garbage-Out criticism (if the data being entered is facetious or not necessarily relevant to the study then the results aren’t necessarily relevant or good) and the conclusions thus far are still relatively generic, two conclusions immediately stood out to me.

The first conclusion that jumped out at me was a very interesting analysis done where the researchers shuffled the precise disease-to-gene relationships while keeping the total number of relationships per gene and per disease the same (and repeated this 10,000 times). The finding from that analysis was that the “network clusters” from a random shuffle tended to be much larger than the actual network cluster size or, in other words, diseases which are genetically related to one another tend to be 8 times more related to one another than one would expect at random. This suggests that there are probably types of disease gene profiles with which most diseases tend to cluster around, something which ties to the groups finding that genes which are “shared” by multiple diseases tend to encode proteins which interact with one another!

The second striking conclusion was that disease-associated genes tend to interact and associate with fewer genes than non-disease associated genes. This is in sharp contrast to diseases like cancer which arise from somatic mutations (mutations which happen after birth and are not passed down from past generations) which almost always affect genes which interact with many other genes. The reasoning the paper gives, while speculative, rings true to me:

“This unexpected peripherality of most disease genes can be best explained by using an evolutionary argument. Mutations in topologically central, widely expressed genes are more likely to result in
severe impairment of normal developmental and/or physiological function, leading to lethality in utero or early extrauterine life and to eventual deletion from the population. Only mutations compatible with survival into  the reproductive years are likely to be maintained in a population.”

I’m sure I’m not the only one who eagerly awaits what other insights can arise from mining large biological databases for network information: the holy grail, of course, being enhancements to the quality of medical diagnoses and treatments.

(Image credit – Diseaseome poster)