Performance of the ForenSeqTM Imagen Kit for Forensic DNA Phenotyping Under Partial Genotyping Conditions

Background: Forensic DNA phenotyping (FDP) enables the inference of externally visible characteristics (EVCs) and biogeographic ancestry when conventional STR profiling is inconclusive. The ForenSeq™ Imagen kit (107 SNPs) integrates phenotype-, ancestry-, and Y-SNPs markers; however, its performance under partial genotyping conditions has not been systematically evaluated. Methods: Ninety-four samples from a Mexican mestizo population were analyzed using the ForenSeq™ Imagen kit on the MiSeq FGx™ platform. Due to incomplete genotype recovery, 41 samples with >60% locus detection were selected for downstream analyses. Phenotype prediction was performed using the HIrisPlex-S model, and ancestry inference was assessed through principal component analysis. In silico simulations were conducted to evaluate locus-specific dropout effects. Results: Eye color prediction showed both reduced feasibility (68.3%) and lower overall accuracy (56.1%), primarily driven by systematic prediction failure when rs12913832 (HERC2) was absent, although accuracy among successfully predicted samples remained high (82.1%). In contrast, hair and skin color inference remained feasible in >97% and 100% of evaluable samples, respectively; however, classification accuracy was moderate (70% for hair and 61% for skin), improving substantially when allowing adjacent-category concordance (90.2% for skin). Ancestry inference was robust when at least 27 aiSNPs were detected, and Y-SNPs reliably distinguished male and female samples. In silico analyses confirmed the critical contribution of rs12913832 to eye color model operability. Conclusions: FDP performance under partial genotyping reflects a trade-off between prediction feasibility and accuracy and depends on locus-specific integrity rather than overall genotype completeness. The ForenSeq™ Imagen kit shows robustness for ancestry, sex, hair, and skin prediction, although with variable accuracy, whereas eye color inference remains structurally vulnerable to drop out of high-impact variants. Evaluating FDP systems under realistic non-ideal conditions is essential to define their true operational limits and ensure scientifically robust and responsible implementation.