STATISTICAL CONFIDENCE FOR LIKELIHOOD-BASED PATERNITY INFERENCE IN NATURAL-POPULATIONS

Paternity inference using highly polymorphic codominant markers is bec oming common in the study of natural populations. However, multiple ma les are often found to be genetically compatible with each offspring t ested, even when the probability of excluding an unrelated male is hig h. While various methods exist for evaluating the likelihood of patern ity of each nonexcluded male, interpreting these likelihoods has hithe rto been difficult, and no method takes account of the incomplete samp ling and error-prone genetic data typical of large-scale studies of na tural systems. We derive likelihood ratios for paternity inference wit h codominant markers taking account of typing error, and define a stat istic Delta for resolving paternity. Using allele frequencies from the study population in question, a simulation program generates criteria for Delta that permit assignment of paternity to the most likely male with a known level of statistical confidence. The simulation takes ac count of the number of candidate males, the proportion of males that a re sampled and gaps and errors in genetic data. We explore the potenti ally confounding effect of relatives and show that the method is robus t to their presence under commonly encountered conditions. The method is demonstrated using genetic data from the intensively studied led de er (Cervus elaphus) population on the island of Rum, Scotland. The Win dows-based computer program, CERVUS dagger, described in this study is available from the authors. CERVUS can be used to calculate allele fr equencies, run simulations and perform parentage analysis using data f rom all types of codominant markers.