Saturday, December 06, 2008
In early 2007, a paper on expression differences between populations claimed that something like 25% of all genes are differentially expressed between two population groups (in that case, in cells lines from people of either European or Chinese origin). That paper, though, had a pretty serious flaw--ancestry effect on expression were completely confounded with microarray batch effects, so the precise numbers in the paper were somewhat suspect.
One way to test whether differences between populations in expression levels are real would be to measure expression on admixed individuals--if expression levels correlate with admixure proportions within a sample, that's pretty good evidence that genetic background plays an important role in expression (barring some third factor that correlates with both genetic background and expression of a large number of genes). A population of admixed European-Asian individuals is probably a little hard to come by, but admixed European-African individuals (AKA African-Americans) are less so. A recent paper lays out the results of a study like this in African-Americans.
The results are somewhat surprising--by correlating expression levels with admixture proportions, the authors speculate that nearly all genes have an ancestry effect on expression. The reason this is somewhat surprising is that, given the way the authors did the analysis, it means the expression of a locus depends on a large number of other loci throughout the genome (if the expression levels of a locus were only affected by variation at that same locus, there would be no correlation between total ancestry and expression). Indeed, the authors estimate that only ~12% of heritable variation in expression of a given gene is due to the effects of local (or cis) variation. Other studies have had little success in identifying distant (or trans-acting) effects in humans, this suggests that the reason, as in many other genome-wide association studies, is simply a lack of power.