Several SNPs have more than one rsid probing the same position on the same chromosome. Concordance of the results from these probes permit an estimate of error rates.

For my own data at these sites, only about half report nucleotide composition. For all but four of these, all probes give the same answer.

Of these discordant four, listed below, two appear to probe opposite strands, one probe reporting CC, the other, GG. The other two sites give incompatible results.

From this, I estimate that 1-3% of the reported SNP results are incorrect.

Is this consistent with other data?

The code to do this analysis can be downloaded from GitHub, at git@github.com:jsh/23andme.git

Summary from my own raw data follows:


rsid           chromosome     position       genotype
------------------------------------------------------------
i4000438       15             70427442       CC
i5004858       15             70427442       GG

rs1135840      22             40852557       CG
i4001403       22             40852557       GG

i4001413       22             40855078       CC
rs61736512     22             40855078       GG

rs28671493     MT             16184          C
i4001306       MT             16184          A


#snps = 966983
indeterminate = 16694
insertions = 850
deletions = 221

duplicates = 402
match = 221
disagree = 4