IBDKinship: An LR-based kinship analysis method for distant relatives up to the seventh degree using IBD segments

Kinship analysis in forensic genetics is essential for establishing biological relationships and has broad applications including paternity testing, missing person identification, and disaster victim identification. Although STR-based methods generally perform well for close relationships such as first- and some second-degree relatives, their resolution decreases substantially for third-degree and more distant relationships. To provide complementary genome-wide evidence for distant kinship analysis, we developed IBDKinship, a novel likelihood ratio (LR) method designed for both kinship inference and testing. The method jointly models the number and length of identity-by-descent (IBD) segments together with their chromosome-specific distribution, enabling improved resolution across a wide range of genetic relationship degrees. We evaluated its performance using simulated pedigree data derived from the 1000 Genomes Project, with variants restricted to the Illumina Global Screening Array (GSA) marker set, covering first- to seventh-degree relationships. For relationship inference, our method achieved high accuracy for close relatives and maintained measurable resolution across distant relationship degrees. Specifically, it achieved 100% accuracy for first-degree relationships, 99.32% for second-degree, and 92.99% for third-degree relatives, followed by 78.29%, 61.68%, 46.05%, and 59.08% for fourth- to seventh-degree relationships, respectively. In kinship testing, where the goal was to distinguish related individuals from unrelated ones, the method demonstrated 100% sensitivity, specificity, and effectiveness for relationships up to the fourth degree, and maintained effectiveness above 99% for fifth- and sixth-degree relationships, with effectiveness of 95.90% for seventh-degree relationships. Overall, this method provides a likelihood-based framework for weighting genome-wide IBD segment evidence, with the strongest support observed for high-quality genotype data and for close to moderately distant relationships.