A new method to detect methylation profiles for forensic body fluid identification combining ARMS-PCR technique and random forest model. (November 2020)

• Record Type: Journal Article
• Authors: Tian, Huan; Bai, Peng; Tan, Yu; Li, Zhilong; Peng, Duo; Xiao, Xiao; Zhao, Huan; Zhou, Yan; Liang, Weibo; Zhang, Lin
• Is Part Of: Forensic science international. Volume 49(2020)
• Publication Date: 2020-11
• Subjects: DNA methylation -- amplification refractory mutation system(ARMS)-PCR -- random forest model -- body fluid identification; Forensic genetics -- Periodicals; Génétique légale -- Périodiques; Forensic genetics; Electronic journals; Periodicals; 614.1
• DOI: 10.1016/j.fsigen.2020.102371
• Languages: English
• ISSNs: 1872-4973

Abstract

Highlights: ARMS-PCR technique could effectively detect the methylation levels in a simple and cost-effective manner. 22 body fluid-specific CpG sites were included in development of the multiplex assays. The random forest model performed well in predicting single source body fluids with high prediction accuracy (99.66%). Random forest model is a practical and promising prediction method for the mixed body fluid identification. Abstract: A set of DNA methylation markers was detected and evaluated to identify body fluids using the amplification refractory mutation system-PCR (ARMS-PCR) and random forest algorithm. In this study, four multiplex DNA methylation reactions composed of 22 promising methylation markers were used to identify regular forensic body fluids, including venous blood, saliva, semen, menstrual blood, and vaginal fluid. The ARMS-specific primers were used to amplify the candidate markers, and then the methylation values of each CpG site were detected through capillary electrophoresis (CE). The DNA methylation patterns of 22 highly informative methylation markers were consistent with previously reported results to a certain extent. To our knowledge, our study is a new method to apply the ARMS-PCR technique and random forest model to detect DNA methylation patterns and identify the type of body fluids in forensic science, thus providing a new method for forensic body fluid identification. Moreover, we proved that this method is robust, applicable and effective for identifying body fluids using the random forest model. The accuracy to predict all body fluids reached up to 0.9966. We firmly believe that this method will have a great potential in the detection of methylation profiles at the molecular level.