We present ibdrel, a user-friendly software for predicting relationships in forensic investigative genetic genealogy (FIGG). In FIGG, an unknown DNA trace is matched against genetic databases, often producing long lists of likely relatives. Accurate relationship predictions for these matches are crucial for guiding subsequent investigative steps, but current methods and software for this are limited. In particular, there is a need for relatedness inference methods handling the complexities of FIGG data. 

Ibdrel implements a probabilistic classification model for relatedness inference, taking as input a list of genomic regions that are shared identical by descent (IBD). A novel feature is that classification can be made at different levels of resolution, from relatedness degree alone to the exact pedigree relationship. This gives users ranked lists, ordered by likelihood, that match their needs. In addition, it incorporates a novel outlier detection tool aimed at detecting inconsistencies between an input and IBD sharing patterns for various relationships. Lastly, many sources operate with a lower threshold on segment length, e.g., 7 cM, and this is incorporated into the classifier.

As part of the theoretical groundwork for ibdrel, we present a classification of indistinguishable pairwise relationships under sex-specific recombination rates. This extends a celebrated result of Donnelly (1983), who assumed equal rates in males and females. The resulting equivalence classes define the finest possible level of classification, and help reduce the workload in ibdrel without sacrificing accuracy.

To show its practical utility, we applied ibdrel to both simulated datasets and real data from recent identification cases. The results suggest that its predictions are more detailed than those of comparable methods, offering a more complete picture for decision makers. The software is open-source and freely available.

Authors

• Henrik Nordtorp (Department of Forensic Sciences, Oslo University Hospital, Norway)
• Magnus Dehli Vigeland (Department of Forensic Sciences, Oslo University Hospital, Norway)