The second example consists of a set of tightly linked missense variants in collagen 6A5 on chromosome 9 at 105.76 Mb (Col6a5, Fig. 4a). Col6a5 is a variant-rich gene and contains 21 missense variants, including a radical substitution (R1778C). Quantitative RT-PCR shows higher expression of the D allele than the B allele in bone marrow (Fig. 4b). As expected, expression differences are strongly associated with the Col6a5 locus in bone expression PheWAS (q=3.5 × 10−4, Fig. 4c). Unlike Fh1, the high density of linked variants in Col6a5 means that we cannot resolve effects of single SNPs, but the scan does define effects of B and D haplotypes for Col6a5 across the phenome. We find that this polymorphic gene is associated with differences in bone mineral density (BMD; GN ID 16532; q=0.037) (Fig. 4d) and quantitative micro-CT analysis confirms a marked difference in cortical BMD at the femoral midshaft between B6 (1069.6±51.4 mgHA per cm3) and D2 parents (1170.8±39.8 mgHA per cm3) (Fig. 4e,f). In humans, mutations in collagen VI are associated with a variety of musculoskeletal abnormalities21. We performed a
