Study on Postmortem Interval Estimation by Proton Nuclear Magnetic Resonance Spectroscopy-based Metabolomics

Cao Jie,Gu Zhen,Liu Ming-Feng,Dang Li-Hong,Du Qiu-Xiang,Li Yu,Sun Jun-Hong

DOI: https://doi.org/10.19756/j.issn.0253-3820.221049

IF: 1.193

2022-01-01

Chinese Journal of Analytical Chemistry

Abstract:Postmortem interval (PMI)estimation is the key issue to be solved in forensic practice. The current technologies and methods lack accuracy,specificity and reproducibility,so reliable and stable molecular markers are needed to provide scientific evidence. In this study,proton nuclear magnetic resonance spectrum (1H NMR)-based metabolomics technology was used to explore the changes of small molecular compounds in rat skeletal muscles at different PMI,and the PMI was predicted combined with machine learning model. Ten Sprague-Dawley rats were sacrificed. Then the skeletal muscle tissues were collected at 0,6,12,18,24,48 and 72 h (n = 10)post-death. The metabolite profiles were obtained by 1H NMR. Twenty molecular markers including hydroxybutyric acid,lactic acid,tyrosine and hypoxanthine were filtered from the metabolite spectrum by orthogonal partial least squares (OPLS)method combined with Mann Whitney U test,database comparison and random forest (RF)feature selection. On this basis,a stacking ensemble model based on RF,gradient boosting decision tree (GBDT),linear discriminant analysis (LDA)as the base classifier and logistic regression (LR)classifier as meta-classifier was established to predict PMI. The prediction performance of the stacking ensemble learning model was better than those of RF,GBDT and LDA. The accuracy of this model for PMI estimation was 85. 71% and the AUC value was 0. 85. The results showed that 1H NMR technique combined with stacking ensemble model could effectively predict PMI by detecting metabolites changes in skeletal muscle of rats at different PMI,providing a new technique and analytical strategy for PMI inference.