There are ~30×106 sites known to be variant in the human population according to dbSNP release 134, which is ~1000 variants/megabase. A given individual typically has ~3×106 variants in their genome, 95% of which fall on dbSNP sites41,42. Therefore we expect ~50 variants/mb not at dbSNP sites, i.e. P(germline| non-dbSNP site) = 5×10−5 and therefore we use θN|non-dbSNP site = 2.2. At dbSNP sites, however, we expect 95% of the ~3×106 variants to occur in the 30×106 sites in the dbSNP database, yielding P(germline| dbSNP site) = 0.095 hence θN|dbSNP site = 5.5.
