DSM distinguishes matching pairs of profiles from mismatching pairs with higher confidence. (A,B) DSM separates most individuals in the test data set from the background distribution of match scores more distinctly than both GNB (A) and EBL (B), as assessed by the P-value of the true match score compared with null distribution of mismatching pairs’ scores (FUSION individuals only). In C, we visualize the strength of separation for the two individuals marked (a and b) in relation to the null distribution normalized as a Gaussian distribution. The DSM indeed separates these individuals more than GNB and EBL. Note that individual a is the one individual that all three methods fail to link, and this individual is substantially less separated from the null distribution than the other individuals.