Kinship assignment with the ForenSeq™ DNA Signature Prep Kit: Sources of error in simulated and real cases

Kinship recognition between anonymous DNA samples is becoming a relevant issue in forensics, more so with the increasing number of DNA profiles in databanks. Also, NGS-based genotyping is being increasingly used in routine personal identification, to simultaneously type large numbers of markers of different kind. In the present work, we explored computationally and experimentally the performance of the ForenSeq™ DNA Signature Prep Kit in identifying the true relationship between two anonymous samples, distinguishing it from other possible relationships. We analyzed with Familias R series of 10,000 pairs with 9 different simulated relationships, corresponding to different degrees of autosomal sharing. For each pair we obtained likelihood ratios for five kinship hypotheses vs. unrelatedness, and used their ranking to identify the preferred relationship. We also typed 21 subjects from two pedigrees, representing from parent-child to 4th cousins relationships. As expected, the power for identifying the true relationship decays in the order of autosomal sharing. Parent-child and full siblings can be robustly identified against other relationships. For half-siblings the chance of reaching a significant conclusion is already small. For more distant relationships the proportion of cases correctly and significantly identified is 10% or less. Bidirectional errors in kinship attribution include the suggestion of relatedness when this does not exist (10-50%), and the suggestion of independence in pairs of individuals more than 4 generations apart (25-60%). The real cases revealed a relevant effect of genotype miscalling at some loci, which could only be partly avoided by modulating the analysis parameters. In conclusion, with the exception of first degree relatives, the kit can be useful to inform additional investigations, but does not usually provide probatory results.