Second, we compute an approximate likelihood ratio score for each potential IBD match identified by the above scan. This procedure is similar in spirit to Parente242, which likewise computes approximate likelihood ratio scores to increase sensitivity and specificity of IBD calls. Our approach prioritizes speed over accuracy; instead of using a haplotype frequency model as in Parente2, we use only allele frequencies and LD Scores43 to compute an approximate likelihood ratio for the observed match having occurred due to IBD versus by chance. We apply this procedure within a seed-and-extend framework in which we begin with long IBS>0 matches but consider extending them beyond IBS=0 sites (to tolerate genotyping errors). We record all extended matches with length >4cM and likelihood ratio >10N (where N is the number of samples) as probable IBD matches.
