Identifying human biological stains detected at crime scenes by comparing DNA profiles to national or international forensic databases is instrumental to provide new orientations to police investigations. When unknown profiles are established, DNA phenotyping offers additional investigation leads by inferring key aspects of individual’s appearance characteristics and ancestry origins. Since 2016, the forensic police laboratory of Lyon (French National Police) has investigated over 80 criminal cases using DNA phenotyping. Until then, we used the SNaPshot method requiring at least 3 ng of DNA to detect both SNPs related to physical traits and INDELs related to genetic ancestry.
In 2026, we implemented the NGS-based method Verogen Forenseq Imagen (Qiagen) to establish SNP profiles for DNA phenotyping using 1ng of DNA or less. This change in ancestry marker type, from INDEL to aSNP (ancestry SNP), led a complete reevaluation of the ancestry prediction pipeline. In addition to an increased reference population size (previously based on 634 individuals) and metapopulation diversity, distinct interpretation methods were introduced, in agreement with recent EDNAP recommandations.
To evaluate aSNP allele frequencies, we collected genomic data from human genetic databases (IGSR, SGDP, HGPD…) and publications. Genetic ancestry was predicted using Snipper app suite version 3, Structure and Genogeographer. SNP profiles were generated for 129 individuals using the Verogen ForenSeq Imagen kit from 1ng of genomic DNA.
We first created the genetic database BAVAR by concatening genetic profiles including up to 1.771.934 SNPs for 6312 individuals with known ancestry.
After quality filtering, we established a reference population for the 56 aSNP from the Verogen Forenseq Imagen genotyping method, based on 6065 genetic profiles from 9 metapopulations : Sub-Saharan Africa, Northern Africa, Native Americans, Eastern Asia, Southern Asia, Middle East, Europe, Oceania and Siberia. We defined additional test genotypes associated to known ancestries (129 individuals genotyped in the laboratory and 1458 individuals from human genetic databases). The clustering of individuals into reference metapopulations was assessed using a leave-one-out approach and Structure. With our reference population, Stucture demonstrated an ancestry prediction accuracy of 89 %, while Snipper and Genogeographer displayed population assignment accuracies of 86 % and 82 %, respectively, for 129 individuals genotyped with the Imagen method. Finally, when combining these three tools, ancestry interpretation accuracy reaches 91 %. This new DNA phenotyping pipeline with improved ancestry preciseness (9 vs 4 metapopulations) will shortly be used by the french National Police for casework investigations.