Of note is the low degree of overlap between HapMap and gene re-sequencing variants. For all 24 candidate genes analyzed, only 12% of variants (179/1550) are shared between available HapMap SNPs and sequencing data generated from our healthy oldest-old. As the LD structure of variants only present in one set cannot be inferred in the other set, HapMap and candidate gene re-sequencing variants have to be treated, for LD-based tagSNP selection, as different data sets. Prioritizing variants found in both sets in our tagSNP selection method, however, increased the number of overlapping SNPs in the tagSNP set to 18% (120/682) ( Figure 1 ).
