| NC_000010.10(GRCh37) || g.96522561T(rs17885098), g.96602623G(rs3758581), g.96522613A>G, g.96609568T>C(rs4917623)|| g.96521422A>G(rs7902257), g.96522613A>G || g.96522613A>G(by liftOver) || Disagree on rs4917623(intron), rs7902257(2kb upstream variant). Disagree on requirement that rs17885098 and rs3758581 must be reference (i.e. only PharmGKB require that these coordinates are not missing). The reason that PharmGKB has included these positions is that they assume different reference bases for these positions (seems like a problem caused by change of Major allele between reference builds GRCh37/GRCh38, which is inverted in PharmGKB vs dbSNP and LiftOver)

| NC_000010.11(GRCh38) || g.94762856A>G || g.94762804C>T(rs17885098), g.94762856A>G, g.94842866A>G(rs3758581) || g.94762856A>G || PharmGKB does not agree with itself when reporting GRCh38 variants in Excel sheets and GRCh37 variants in the API. The differences probably caused by non-standard use of Major/Minor Allele (rs17885098, rs3758581), causing reference bases with respect to be reported as variants for GRCh37 and not GRCh38, as would be expected given info from dbSNP. Filtering out intron variants (rs4917623) in the Excel sheet may be sensible from an exon/protein-coding view.