Large-scale genetic association studies are now widely conducted using SNPs selected from the International HapMap Project or provided on commercial "whole genome" chips. As only a subset of human genetic variation has been identified, it is unknown what proportion of the total genetic variation can be captured in this way, although recent genome-wide estimates of SNP capture rates have been encouraging. We estimated the expected gene-centric information capture for whole-genome chips using sequence data from 306 inflammatory/cardiovascular genes and found SNP capture rates notably lower than previous genome-wide estimates. Further investigation indicates that a major explanation for these lower capture rates is the aggregation of particular sequence features that influence both linkage disequilibrium and the amenability of SNPs for genotyping within the broad class of inflammatory/ cardiovascular genes. This suggests that the power of genetic association studies in some complex traits will depend not only upon established factors, such as allele frequency and penetrance, but may also be influenced by the distribution of sequence features in the class of genes expected to underlie the disease of interest.