混合痕跡からのバルクトランスクリプトームデータを用いた体液提供者の同定

The deconvolution of the mixture contributors is essential but always presents challenges in crime scene investigation, especially when encountering mixtures with multiple contributors. In recent years, RNA-based genotyping has shown great advantages in assigning a body fluid to a specific mixture contributor, whereas transcriptome sequencing has been seldom applied for mixture deconvolution. In this study, we investigated the feasibility of bulk transcriptomes data for deconvoluting biological mixtures, and described a novel approach, utilizing bulk transcriptomes data of multi-body-fluid mixtures to identify body fluid donors. Computational deconvolution methods were introduced here to infer body fluid proportions from bulk transcriptomes data, and SNPs corresponding to a specific body fluid donor could be separated from differentially expressed genes with the knowledge of body fluid composition. The described method was tested on both in-silico and real mixtures. After a quantitative evaluation, robust regression-based deconvolution methods with unnormalized expression profiles were finally applied to body fluid deconvolution tasks, and body fluid compositions of mixtures could be determined with high accuracy according to the deconvolution results. Finally, the corresponding SNPs were successfully extracted from both in-silico and real mixtures, producing high likelihood ratios and strong evidentiary weight for individually identifying body fluid donors. Altogether, our proof-of-concept study established the utility of bulk transcriptomes data to identify mixture contributors, providing new avenues for mixture deconvolution.