Developmental validation of a microRNA panel using quadratic discriminant analysis for the classification of seven forensically relevant body fluids. (July 2022)

• Record Type: Journal Article
• Authors: Rhodes, Ciara; Lewis, Carolyn; Szekely, Jennifer; Campbell, Annabelle; Creighton, Mary-Randall A.; Boone, Edward; Seashols-Williams, Sarah
• Is Part Of: Forensic science international. Volume 59(2022)
• Publication Date: 2022-07
• Subjects: Forensic body fluid identification -- MicroRNA and miRNA -- Developmental validation -- Reverse transcription-quantitative polymerase chain reaction and RT-qPCR -- Quadratic discriminant analysis and QDA; Forensic genetics -- Periodicals; Génétique légale -- Périodiques; Forensic genetics; Electronic journals; Periodicals; 614.1
• DOI: 10.1016/j.fsigen.2022.102692
• Languages: English
• ISSNs: 1872-4973

Abstract

Abstract: Body fluid identification is an important step in the forensic DNA workflow, and more advanced methods, such as microRNA (miRNA) analysis, have been research topics within the community over the last few decades. We previously reported a reverse transcription-quantitative PCR (RT-qPCR) panel of eight miRNAs that could classify blood, menstrual secretions, feces, urine, saliva, semen, and vaginal secretions through analysis of differential gene expression. The purpose of this project was to evaluate this panel in a larger population size, develop a more statistically robust analysis method and perform a series of developmental validation studies. Each of the eight miRNA markers was analyzed in > 40 donors each of blood, menstrual secretions, feces, urine, saliva, semen, and vaginal secretions. A 10-fold cross-validated quadratic discriminant analysis (QDA) model yielded the highest classification accuracy of 93% after eliminating miR-26b and miR-1246 from the panel. Accuracy of body fluid predictions was between 84% and 100% when various population demographics and samples from the same donor over multiple time periods were evaluated, but the assay demonstrated limited scope and reduced accuracy when mixed body fluid samples were tested. Limit of detection was found to be less than 10 4 copies/µL across multiple commercially available RT-qPCR analysis methods. These data suggest that miR-200b, miR-320c, miR-10b, and miR-891a, when normalized to let-7 g and let-7i, can consistently and robustly classify blood, feces and urine, but additional work is important to improve classification of saliva, semen, and female intimate secretions before implementation in forensic casework. Highlights: A novel panel of six miRNAs was developed to classify seven body fluids. QDA analysis of over 500 samples resulted in an overall prediction accuracy of 93%. The QDA model is publicly available for testing at https://tinyurl.com/y9e5c9ca . Using the current QDA model, mixtures were difficult to predict. Alternate detection or analysis methods may be necessary to deconvolute mixtures.