To minimize the number of variants that have to be genotyped in association studies, we devised an innovative tagSNP selection protocol outlined in Figure 1 . This protocol is designed to incorporate both common and less common variants. Because variants discovered in healthy oldest-old represent potential healthy aging alleles, we include SNPs with MAF ≥2% that were found in re-sequencing. This was equivalent to including variants that had been observed at least twice, but excluded ‘singletons’ that had been observed only once. To pick tagSNPs representing SNPs found by re-sequencing, we set a high threshold of r2 = 1.0, since we did not want to leave out any SNPs from this unique source. To cover genetic variation in the introns and gene proximity we also included HapMap SNPs for the European ancestry population (CEU) chosen based on genomic regions of candidate genes±10 Kb. MAF ≥5% and r2 = 0.8 were used to choose representative tagSNPs from among these known SNPs.
