Resolving kinship degrees with identity-by-descent coefficients: Limits and insights from a simulation study

Abstract

Kinship analysis in forensic genetics often requires precise determination of relatedness degrees rather than existence of kinship. While identity-by-descent (IBD) coefficients are widely used, their discriminatory power for distinguishing relatedness degrees under a likelihood ratio (LR) framework remains inadequately characterized, particularly for distant relatives. In this study, we developed a maximum LR approach with dynamic thresholds to evaluate IBD-based kinship discrimination. LR_kin calculations were tested using competing hypotheses (H0: relationship i vs. H1: relationship j, i ≠ j). System performance was assessed across seven virtual panels: five STR panels (21–500 loci) and two SNP panels (6445–16,109 loci), simulating panels with different effectiveness. The results showed that the LR_kin-based maximum LR method effectively differentiated closer relationships, such as parent-offspring, full-sibling, and second-degree relatives. However, the LR_kin struggled to distinguish more distant relationships, even using up to 500 STRs or 16,109 SNPs. In conclusion, although calculating LR_kin does not improve inference accuracy, it can help reduce the risk of errors in inferring relationships. Using more genetic markers can enhance inference accuracy; However, the IBD-coefficients based LR had limitations in distinguishing between different degrees of relationships.