Of mice and men and meta analyses
| 31 August, 2011 | Richard P. Grant |
|
|
Q: When is a gene not a gene?
A: When it’s a paralogue.
As a joke, that could probably use some work. But it touches the core of genome annotation, and how we can estimate the functions of genes in non-model organisms (i.e. in organisms where it is difficult if not impossible to directly determine the function).
Orthologous genes are those in different species that share a single common ancestral gene in the last common ancestor of both species. Paralogous genes are those which have duplicated and diverged within the same vertical branch of the phylogeny. And an implicit tenet of genome annotation is that orthologous genes typically have the same, or at least closely related, functions.
So a paper in PLoS Comput Biol challenging this ‘ortholog conjecture‘ (free F1000 evaluation) looks set to create quite a stink, and might bring into question the “whole foundation of modern genome biology”–at least, if the conclusions of the paper are true. Which Michael Galperin of the NIH disputes in a dissent. Michael’s argument is that the paper’s authors failed to “painstakingly” compare “functions of orthologs in two reasonably well characterized genomes,” and instead compared Gene Ontology (GO) annotations for orthologous and paralogous genes. A meta-analysis, if you will–or to be harsh about it, studying what others say about a problem–gene function in this case–rather than the problem itself.
The problem with the paper is that GO annotations in mice and men are not independent, and the author’s technique is more likely to uncover a bias in annotation than any real biological effect. Michael himself tried, and failed, to find examples where orthologous genes in men and mice have been experimentally characterized, and shown to have distinct functions. As a result, he contends that (fortunately), the paper’s “conclusions are wrong, and the key principle of comparative genomics is safe for now.”
|
|
previous post
I also have my doubts about the findings of this paper. One important thing to note is that their analysis has only been carried out between a single pair of organisms. They only look and GO annotations and, in a supplementary manner, expression data. That being said I also have issues with Mr. Galperin’s analysis of the paper.
First off, I am curious exactly how experimental annotations in mice and men not being independent would serve to refute the “ortholog conjecture” more than it would support it. One could just as easily argue that this dependence would bias annotations more towards orthologs being functionally similar. Furthermore, where is Mr. Galperin’s evidence that this is even the case (the dependence of human and mice annotations), or was this just the result of his own meta-analysis of the situation? Does he have a citation for this statement?
Furthermore, I’d like to know what his definition of and “experimentally well characterized” organism is. Human and Mouse are very well characterized, and are also, amongst organisms that are well characterized, the pair that share the most recent time of divergence.
He also dismisses their findings as merely “meta-analysis”. Even if you consider it so, they are directly comparing experimental GO annotations, an important source of information used to characterize genes. I believe if they were just studying “what others say about the problem” they would have ended their paper half way through the second page where they note there has never been a large-scale analysis of the ortholog conjecture.
Finally, and most importantly, I find the assumption that all of comparative genomics would be useless if their findings did hold to be a little extreme. Note that the paper only finds that on average paralogs are more functionally similar than orthologs. It also seems that they find that orthologs are still more functionally similar than what you would find by random. Phylogenetic methods would still be extremely important, and using one organism to make inferences about the biology of another would certainly still be possible if “the ortholog conjecture” didn’t hold.
Whether what they find folds or not, it seems this will be a win-win for science. The community will be spurred to look more closely at GO annotations in order to determine whether there is bias in how they are curated. If no bias is found, then will have updated and elaborated our basic understanding of evolution.