A single-cell transcriptomic method for the deconvolution of multi-donor blood mixtures

Identifying individual donors from multi-donor biological mixtures remains challenging in forensic science, especially for same-tissue samples. Single-cell RNA sequencing (scRNA-seq) preserves individual genetic variation through cellular isolation, offering a promising solution. However, there is a lack of systematic evaluation regarding procedures for analyzing forensic mixtures. Here, we present and systematically evaluate a donor-reference-free deconvolution framework across complex three-donor and five-donor peripheral blood mixtures. To overcome transcriptomic sparsity and technogenic noise, we implemented a two-stage filtering strategy combining population-level pseudo-bulk variant prescreening via FreeBayes with cell-level genotyping using CellSNP-lite. This approach effectively purges false-positive candidate single-nucleotide polymorphisms, doubling the single-cell matrix density and increasing high-depth loci by nearly seven-fold. When integrated with the Vireo algorithm, the optimized upstream data quality elevates downstream genetic divergence scores, enabling clear individual deconvolution. Independent whole-genome sequencing validated the framework, demonstrating high genotype concordance and generating powerful likelihood ratios that far exceed conventional forensic identification thresholds. Furthermore, the deconvolved single-cell profiles enable reliable cross-sample donor matching and demonstrate concordant biogeographic ancestry assignments consistent with the known East Asian background of all participating contributors. This proof-of-concept study establishes a validated, practical framework for resolving same-tissue blood mixtures, providing a foundational baseline for the application of single-cell transcriptomics in forensic genetics.