I notice the GRanges returned by the dbSNP packages have strand '*'. Does anyone know how safe am I in assuming that the variant alleles also given by the package actually correspond to the '+' strand? This seems to be the case for the 20 or so I have checked manually, but maybe I have just been lucky. I ask this in the context of trying to use predictCoding in the VariantAnnotations package to find coding SNPs. For SNPs in genes on the '-' strand I have found that I have to complement the alleles given by dbSNP to get the correct result. I just want to make sure that assuming the alleles are from the '+' strand is a reasonable assumption in the vast majority (say >99%) of cases. I realise from my reading of the SNPlocs.Hsapiens.dbSNP.20110815 manual that some SNPs will be incorrect anyway (it mentions ~0.1% of SNPs not mapping to the reference at all), that level of failure is acceptable, but anything higher would be a worry. -- Alex Gutteridge