Benchmarking KinSNP®: A study on genetic relationship prediction for forensic applications

Area of Science:

• Forensic Genetics
• Human Identification
• Computational Biology

Background:

• Accurate kinship inference is crucial in forensic DNA analysis for unidentified remains and compromised samples.
• Existing methods often rely on Short Tandem Repeat (STR) profiles, which may not always be obtainable or have reliable references.

Purpose of the Study:

• To evaluate the performance of KinSNP®, a forensic software tool, for predicting genetic relationships using whole genome single nucleotide polymorphism (SNP) data.
• To benchmark KinSNP® against established kinship analysis tools using real pedigree data.

Main Methods:

• A cohort of 12 individuals with known relationships up to the 6th degree was used for 66 pairwise comparisons.
• KinSNP® was benchmarked against GEDmatch PRO™ and the Shared cM Project 4.0 tool using shared centimorgan (cM) thresholds of 7 and 12 cM.
• Performance was assessed based on correct relationship identification and the occurrence of false positives.

Main Results:

• KinSNP® correctly identified 82% of known relationships in the highest predicted category and 12% in the second highest.
• Predictive confidence decreased for distant relationships (≥5th degree), with spurious matches observed at lower cM thresholds.
• Increasing the cM threshold to 12 cM reduced false positives without impacting sensitivity for close relationships.

Conclusions:

• KinSNP® demonstrates comparable accuracy to existing tools for SNP-based kinship analysis in forensic contexts.
• The software, part of the secure SNPSuite application, offers a reliable offline solution for disaster victim identification and missing person investigations.
• Findings support KinSNP®'s utility when STR profiling is not feasible or direct reference samples are unavailable.