Saturday, November 04, 2006

Accelerated human evolution in non-coding regions? (reprise)   posted by p-ter @ 11/04/2006 03:37:00 PM

The other day, I mentioned a paper in Science claiming that non-coding regions have played a large role in human evolution. In the paper, the authors devise a new test for finding human-lineage-specific evolution, and apply to to a number of conserved non-coding regions in the genome. The new method seems pretty useful--it takes into account both locus-specific neutral evolution rates and lineage-specific constraint. As I mentioned before, they find 992 regions in the genome with human-specific acceleration. But how many would they expect to find by chance? An important question, and one they answer in the Supplementary Materials online:
By the definition of P-values, the expected number of CNSs in any P-value bin of width w is w*110,549 under the null model of constrained human lineage evolution (w = 0.00125 in Figure 1). Thus, we expect only 553 human-accelerated CNSs at a P-value threshold of 0.005 (0.5%), though we observe 992 (0.9%). The false-positive rate in the set of predicted accelerated CNSs is therefore ~56% (553/992).
Hey now! That's a pretty serious false positive rate. So to revise my previous thoughts, they definitely show some non-coding sequences have undergone a human-lineage-specific acceleration. Other groups are getting similar results, so I'm not going to doubt that. But that false positive rate seriously calls into question their conclusion that the accelerated regions are preferentially found hear genes involved in neuronal adhesion--it's tough to argue about the characteristics of a set of regions if more than half of them don't belong in the set at all.

Further, and this is a question that has been bothering me about a lot of papers on human evolution, where do you go from here? Now they have a set of 992 candidate regions that could play an important role in human evolution, but more than half of them are false positives--how do you decide which ones to follow up? It's almost enough to make a guy consider doing "real", wet biology...