EMPIRICAL DETECTION OF NATURAL SELECTION AT TWO BRAIN-RELATED GENES (FOXP2 AND AHI1) BUT FAILURE TO CONFIRM EVIDENCE AT ASPM. F. Yu1, 2, R.S. Hill2, 3, 4, A.A. Mignault1, R.J. Ferland3,4, C.A. Walsh2,3,4, D. Reich1,2. 1) Dept Genetics, Harvard Med. School; 2) Broad Inst. of MIT & Harvard; 3) Div. of Neurogenetics & Howard Hughes Med. Inst., Beth Israel Deaconess; 4) Dept. of Neurology, Harvard Med. School.
A startling hypothesis is that genes cause neurodevelopmental defects have been subject to natural selection during human evolution. Dorus (2004) showed that this has been the case over the past tens of millions of years of primate history, and anecdotal studies have suggested that the selection may have continued into the last tens of thousands of years. The most striking evidence has been at FOXP2 (Enard 2002), ASPM (Mekel-Bobrov 2005), and Microcephalin (Evans 2005). However, the evidence is based on comparing the pattern of variation at the genes to computer simulations, which cannot capture all features of real data. To empirically assess whether the brain-related genes indeed stand out compared with random genomic regions, we studied two genes with previous evidence of selection, FOXP2 and ASPM, and two that cause neurodevelopment defects but with no previous evidence, AHI1 and GPR56. We resequenced 15-30 kb from each in 16 European Americans and 16 West Africans, and genotyped the SNPs in HapMap samples. This was identical as method used to study 5 ENCODE regions, providing a large empirical comparison data set. Both FOXP2 and the AHI1 stand out compared with the pattern of variation in European Americans in the ENCODE regions. The results at AHI1 constitute new evidence and are particularly striking, with the pattern more extreme than anything seen in comparable segments of the ENCODE regions by various tests(all P<<0.006). There is no evidence for selection ASPM, even though a recent study suggested positive selection within the last 6,000 years. Our analysis suggests it arose tens of thousands of years ago, and the gene does not stand out from empirical data by multiple tests. Thus, comparison to simulations can mislead inferences. To provide the most compelling arguments for selection, multiple lines of evidence should be brought to bear, and comparison to empirical data can be an important resource.