Currently I am analyzing SNPs from different genome versions and chips (1kg, 23andme, ALFRED...). I have stuck to the policy that as long as the SNPs have the same rs number, chromosome, position and alleles they are the same across studies and that mixing them can't introduce biases/problems.
Is this a relatively safe assumption or are there still pitfalls?
•
written
by
Endre Bakken Stovner
Different genome versions will have different position offsets. You could validate older genome versions by using a tool like LiftOver on the positions then seeing if the new coordinates agree with the latest genome dbSNP coordinates.
Comparing across genome versions always introduces errors and edge cases.
Note from NCBI: "Note that SNP experiments might be performed at a later time, and possibly contributed by labs other than the one who provided the original submission." I would assume that rs# SNPs can evolve over time with new data as dbSNP is built, e.g. a G/A could become a G/W with more data at the same location.
Log in to answer this question.
