法医学杂志2026，Vol.42Issue(2)：87-93,7.DOI:10.12116/j.issn.1004-5619.2025.450505
联合肠道特征菌与机器学习的裸鼠死亡时间分段推断模型构建
Construction of a Segmented PMI Estimation Model Integrating Intestinal Micro-bial Signatures and Machine Learning in Nude Mice
摘要
Abstract
Objective To observe stage-specific changes in the intestinal microbiota of nude mice after death and to develop a postmortem interval(PMI)estimation model based on"rupture points",thereby exploring a new model for PMI estimation.Methods A total of 108 nude mice were sacri-ficed,and cecal contents were collected at 18 time points(0,24,41,48,55,65,72,79,89,96,103,113,120,144,168,192,216,and 240 h postmortem).16S rRNA gene amplicon sequencing was used to analyze the changes in intestinal microbiota.Based on microbial abundance,a random forest model was employed for cross-validation to identify signature bacterial genera.A segmented regression model was then constructed to estimate PMI and compared with a direct regression model.Results Both α-diversity and β-diversity analyses indicated significant changes in the relative abundance of intestinal microbiota during the periods of 0-103 h and 113-240 h postmortem in nude mice.The segmented re-gression model built using the random forest algorithm achieved an R2 of 0.96 and a mean absolute error(MAE)of 9.83 h for PMI estimation.In contrast,the direct regression model yielded an R2 of 0.81 and an MAE of 16.91 h.Conclusion Microbial succession during cadaver decomposition exhibits clear temporal and stage-specific characteristics.A segmented regression model for PMI estimation using"rupture points"can improve the accuracy of PMI estimation in nude mice.关键词
法医病理学/死亡时间推断/肠道菌群/机器学习/分段回归模型/裸鼠Key words
forensic pathology/postmortem interval estimation/intestinal microbiota/machine learning/segmented regression model/nude mice分类
医药卫生引用本文复制引用
张祥彦,杨帆,胡胜,贾琼,聂昊,赵兴春,郭亚东..联合肠道特征菌与机器学习的裸鼠死亡时间分段推断模型构建[J].法医学杂志,2026,42(2):87-93,7.基金项目
国家自然科学基金面上项目(82471922) （82471922）
公安部科技强警基础工作专项(2023JC17) （2023JC17）
法医病理学公安部重点实验室开放课题资助项目(GAFYBL202401) （GAFYBL202401）
