Related Experiment Video
Updated: Jun 18, 2026
12:19
Rapid Subtractive Patterning of Live Cell Layers with a Microfluidic Probe
Published on: September 15, 2016
ML-Powered MultiFluid-mRNAplex31: A Self-Developed Panel for Precise Tracing of Five Body Fluids
Meiming Cai1, Qiong Lan1, Man Chen2
1Guangzhou Key Laboratory of Forensic Multi-Omics for Precision Identification, School of Forensic Medicine, Southern Medical University, Guangzhou 510515, China.
Genomics, Proteomics & Bioinformatics
|June 16, 2026
Summary
Forensic scientists can now precisely identify single-source and mixed body fluids using a new mRNA multiplex system and machine learning. This advanced method improves accuracy in crime scene investigations by reliably distinguishing between blood, saliva, and other crucial biological samples.
Area of Science:
- Forensic molecular biology
- Biotechnology
- Genomics
Background:
- Messenger RNA (mRNA) expression profiling is a sensitive method for identifying body fluid origins.
- Current methods face challenges like false positives due to marker cross-reactivity.
- Accurate body fluid identification is crucial for linking individuals to criminal activity.
Purpose of the Study:
- To develop an advanced mRNA multiplex amplification system for precise body fluid identification.
- To create robust machine-learning models for analyzing both single-source and mixed body fluid samples.
- To enhance forensic capabilities in characterizing crime scenes through reliable biological sample analysis.
Main Methods:
- Developed MultiFluid-mRNAplex31, an mRNA multiplex system with 26 body fluid-specific mRNAs, 2 sex-specific markers, and 3 housekeeping genes.
- Utilized reverse transcription-polymerase chain reaction (RT-PCR) and capillary electrophoresis (CE) for multiplex amplification detection.
- Created and evaluated four machine-learning models (random forest, SVM, NN, CatBoost) for body fluid classification.
Main Results:
- The MultiFluid-mRNAplex31 system demonstrated exceptional species specificity, sensitivity, and robustness with aged and mixed samples.
- Machine-learning models accurately identified five types of body fluids in single-source samples.
- Support Vector Machine (SVM) and Neural Network (NN) models achieved an average accuracy exceeding 0.96 for mixed samples after 1000 validation iterations.
Conclusions:
- The developed system provides a comprehensive analysis for precise body fluid identification in forensic science.
- This technology offers strong technical capabilities for distinguishing common single-source and mixed body fluids.
- The integration of mRNA profiling and machine learning significantly advances forensic molecular biology.
