Multilocus DNA fingerprinting: the independence problem in quantitative paternity testing.

A simulation study was performed in order to determine whether or not the assumption of independence, made in the quantitative analysis of multilocus DNA fingerprints, represents an inadmissible over-simplification. A total of 10,000 cases of true and false paternity, respectively, were simulated in twenty replicas of various genetic models. Log-likelihood ratios (paternity vs. non-paternity; LR) were calculated using published likelihood formulae and assuming position-wise independence. The resulting LR distributions were compared to (i) the results of a classical analysis of the underlying genotype data, and (ii) the distributions expected from the likelihood model employed in the LR calculations. Although considerable discrepancies were observed between these distributions, decision making about paternity appeared to be only marginally affected, especially when only a fraction of each multilocus DNA fingerprint was analyzed.