Identification of body fluid sources based on microbiome antibiotic resistance genes using high-throughput qPCR

Abstract

Identifying the origin of body fluids is a critical step in forensic investigation. Recently, the development of high-throughput sequencing technology has led to the use of microbiomes for body fluid identification in forensic studies. However, high-throughput sequencing data are difficult to analyze, the sequencing protocol is complicated. An increasing number of studies have focused on antibiotic resistance genes (ARGs) in the human microbiome. The abundance and diversity of ARGs in different parts of the human body can be detected using quantitative polymerase chain reaction (qPCR). To date, no studies have inferred the sources of body fluids based on ARGs. Therefore, we attempted to use ARGs as a tool to infer the origin of body fluids. We assessed the abundance and diversity of 64 ARGs in blood, semen, saliva, vaginal secretions (VS), nasal secretions (NS), and fecal samples using high-throughput qPCR. The results showed that ARGs were more diverse in fecal samples, which was significantly higher than those of other sample types (P < 0.05). Principal coordinate analysis (PCoA) showed that the samples clustered mainly according to their type. We constructed a random forest classification model based on 64 ARGs with a prediction accuracy of 92.68 %. Next, we evaluated the importance of the features in the random forest model (mean decrease accuracy, MDA). Subsequently, we constructed prediction models for the top 40 and 20 ARGs after sorting genes with the highest MDA, and their prediction accuracies were both 92.68 %. The accuracy of the top 10 ARGs was 87.80 %. Notably, when only the top 10 characterized ARGs were used to construct models for saliva, semen, and VS samples, the prediction accuracy reached was 95.24 %. This shows that blood, semen, saliva, NS, VS, and fecal samples can be accurately identified using ARGs. Our results suggest that ARGs are promising markers for forensic body fluid identification.