Single nucleotide polymorphisms (SNPs) offer major advantages for forensic kinship analysis, particularly for degraded samples and more distant relationships. However, routine casework adoption remains limited by the cost, turnaround time, and infrastructure demands of microarray and next-generation sequencing platforms. In addition, many published SNP panel studies are largely simulation-based, use small pedigrees, or assess limited populations, reducing operational relevance. This study aimed to identify an optimised subset of informative SNPs suitable for a reduced and more practical forensic kinship panel.
An initial set of >650,000 fixed markers from Illumina’s Global Screening Array (GSA) v3 was processed through a multi-stage filtering pipeline. Ensembl Variant Effect Predictor was used to prioritise non-coding loci. Multiallelic variants were removed, and the remaining biallelic SNPs were filtered using ALFA allele frequency (AF) database with South Asian continental population specific AFs. Linkage disequilibrium pruning was performed using SNPclip (LDlink) to generate progressively reduced panels under varying stringency thresholds. Filtered SNP sets were evaluated in a multigenerational South Asian pedigree consisting of 35 individuals, extending to fourth-degree relationships. Pairwise likelihood ratios (LRs), relationship indices (RIs), and per-locus informativeness were calculated in R via pedigree-based statistical frameworks (pedsuite).
Preliminary analyses demonstrate that substantial reduction of the initial SNP set can be achieved while retaining strong discriminatory power for kinship inference. First-degree relationships were consistently resolved with high confidence using relatively small SNP subsets. Second-degree relationships required larger panels but remained distinguishable under moderate LD pruning thresholds. Third- and fourth-degrees showed greater sensitivity to marker number and allele frequency distribution, highlighting the importance of population-informed SNP selection. Ranking loci by per-marker LR contribution enabled identification of highly informative markers and estimation of the minimum SNP set required to achieve predefined evidential thresholds (e.g., LR ≥ 1000). Preliminary findings suggest that panels comprising a few hundred carefully selected SNPs may be sufficient for robust close- and mid-range kinship analysis.
This study demonstrates a practical route from high-density SNP discovery to reduced forensic panels with realistic casework potential. By integrating functional filtering, population-aware allele frequency selection, LD pruning, and pedigree-based statistical evaluation, the approach supports the development of smaller, operationally relevant SNP assays for kinship testing. At this stage, panel design has focused on SNPs alone. Future work will assess performance alongside STR markers, with the aim of developing an even smaller and more affordable SNP panel to complement existing STR workflows and improve resolution in challenging kinship cases.