Ancestry-informative microhaplotypes (MHs) were proposed as an alternative to single nucleotide polymorphisms (SNPs) for biogeographical analysis. However, estimation of biogeographic background from MHs is not supported by current assignment software (e.g., Snipper and GenoGeographer), and there is insufficient population data for MHs to allow robust haplotype frequency estimates. In this study, the biogeographical assignment performance of the MHappaMundi MH ancestry panel was evaluated using decomposable graphical models to account for allelic dependence within MH data and estimate haplotype frequencies for all possible haplotypes, including rare and unobserved ones. MHappaMundi data were collected for eight metapopulations (representing 33 subpopulations). Population assignment of MH profiles consisted of an outlier detection test for each reference population prior to calculation of evidential weights to prevent misassignment of individuals whose origin population is unrepresented in the database. Comparison with SNP-based assignments from the Precision ID Ancestry Panel for the same individuals was performed at inter- and intra-continental levels. Fewer MH profiles than SNP profiles were rejected in all populations using the outlier tests. Concordant assignment rates were similar for both MH (94.9%) and SNP (94.7%) data in metapopulations, however, MHs achieved higher concordance rates for subpopulations compared to SNPs (80.2% vs. 54.3%). A total of 4,100,000 MH profiles were simulated to estimate the expected distribution of likelihood ratio (LR) values and theoretical confidence intervals for each combination of populations, providing a basis for the definition of LR thresholds for population assignment. Our results demonstrated that MHappaMundi outperformed the Precision ID Ancestry Panel for population assignment.
Forensic Investigative Genetic Genealogy (FIGG) has emerged as a powerful tool for solving cold cases through the analysis of dense SNP data. FIGG methodologies provide higher sensitivity by detecting a larger fraction of the genome that relatives share compared to standard analysis. This study evaluates the suitability of the Axiom™ Spain Biobank Array (Thermo Fisher Scientific, TFS), which consists of 756,834 SNPs, and it is typically used in biomedical research, for distant kinship inference and its performance in forensic samples. Therefore, considering the commonly encountered incomplete profiles and the high number of SNPs needed to solve distant kinship cases with FIGG approaches, we explored the performance of genotype imputation and refinement in low-quality samples and its impact on kinship analysis. Given the high SNP density of the Axiom array, we assessed whether it could also yield forensic leads regarding phenotype prediction, biogeographical ancestry, and mitochondrial and Y haplotypes. A total of 91 samples were analysed, including samples from i) individuals with known kinship relationships spanning from direct to 7th degree; ii) individuals of diverse biogeographical ancestries and phenotypes; iii) bone samples of varying post-mortem intervals, and iv) two DNA controls. Kinship analysis was performed using the KING approach, segment analysis and likelihood analysis. While the KING approach failed to distinguish relationships beyond third degree, the combination of segment and likelihood analyses successfully resolved most of the distant kinship cases, from third to seventh degree (with the exception of 2 out of 29 cases). Imputation of simulated low-quality datasets resulted in a small error rate, with accuracy decreasing as SNP input declined. Imputation improved performance in segment-based analyses, which require high SNP density to detect shared genomic regions. Genotype refinement was performed in challenging forensic samples aiming to determine the reliability of the obtained genotypes. However, considering that degraded samples produced unreliable genotypes, and refinement did not significantly improve data quality, the array is not suitable for the analysis of skeletal remains and low-quality samples. The SNP array also enabled inference of biogeographical ancestry and mitochondrial and Y-chromosome haplotypes, whereas prediction of externally visible characteristics showed lower accuracy. These findings highlight both the potential and limitations of high-density SNP arrays for FIGG in forensic contexts, particularly when working with compromised DNA.
The targeted analysis of diverse DNA markers, including single nucleotide polymorphisms (SNPs), insertion deletion polymorphisms (InDels) and DNA methylation levels, is central to forensic DNA phenotyping, enabling inference of BGA, appearance traits, and age. Such information is relevant when DNA comparison after conventional DNA profiling fails to identify a person of interest. Recent technological developments allow both genetic and epigenetic markers to be analyzed in long-reads within a single platform such as Oxford Nanopore Technologies (ONT). This approach offers an advantage over traditional capillary electrophoresis and targeted, short-read massively parallel sequencing (MPS) methods by allowing genetic, and epigenetic marker detection in one assay. These (epi)genetic markers for DNA phenotyping have previously been evaluated using the ONT MinION platform, which may improve upon using the PromethION device with a larger pore count which we evaluated in this study. We included not only appearance and ancestry prediction, but also lineage and individual identification. Both whole-genome sequencing (WGS) and adaptive sampling were applied in which STRs, SNPs, InDels, and whole mtDNA were analyzed under both standard (600 ng) and lower-input (<100 ng) conditions. Subsequently, FDSTools, a bioinformatic tool designed for MPS-based forensic marker analysis, and Clair3, a tool for identifying genetic variants in long-read data, were assessed for their suitability in processing nanopore sequencing data. STR allele calling using FDSTools with an allelic balance threshold of 0.25 achieved the highest performance, with F1-scores corresponding to accuracies of 89.1–99.2% across the experimental setups. For InDels, FDSTools also performed best, yielding an F1-score of 98.1%. SNP analysis of 1231 autosomal, X-, and Y-chromosomal markers achieved F1-scores of 98.1–99.8% after applying additional quality thresholds. MtDNA analysis demonstrated F1-scores of 85.9–88.6% for whole mtDNA, and 100% for the control region. Overall, adaptive sampling with standard DNA input produced the highest precision, recall, and F1-scores across all variant types, followed by WGS and lower-input adaptive sampling. While slight variation in appearance probabilities and ancestry proportions were observed due to low level sequencing errors, no differences in appearance and ancestry conclusions between ONT and reference data were seen. Our findings demonstrate the potential of the PromethION platform for integrated forensic DNA analysis by enabling simultaneous analysis of numerous genetic markers of different types for different forensic purposes.
