A preliminary study on cause‑of‑death discrimination and the pathological stage identification in acute ischemia...

Area of Science:

• Forensic Medicine
• Biochemistry
• Computational Biology

Background:

• Accurate diagnosis of acute ischemia heart disease (AIHD) and its pathological staging remain significant challenges in forensic medicine.
• Existing diagnostic methods for AIHD require improvement, highlighting the need for novel biomarkers with enhanced sensitivity and specificity.
• Post-mortem variations in lipid profiles show promise for identifying causes of death, including AIHD.

Purpose of the Study:

• To systematically analyze the non-targeted lipid metabolism profile in post-mortem blood samples from AIHD and non-cardiac disease death cases.
• To identify specific lipid metabolites that can discriminate AIHD and its pathological stages (early myocardial ischemia and acute myocardial infarction).
• To evaluate the performance of machine learning algorithms in classifying AIHD based on lipidomic data.

Main Methods:

• Ultra-high performance liquid chromatography-mass spectrometry (UHPLC-MS/MS) was employed for comprehensive lipidomic profiling.
• A total of 665 lipid metabolites were detected and analyzed.
• Eight machine learning algorithms, including eXtreme Gradient Boosting (XGB) and Logistic Regression (LR), were utilized for classification model development.

Main Results:

• 18 lipid metabolites were identified as key discriminators for AIHD.
• 47 lipid metabolites were found to be important for identifying early myocardial ischemia (EMI) and acute myocardial infarction (AMI).
• Optimized XGB model achieved an AUC of 0.830 and accuracy of 0.781; optimized LR model achieved an AUC of 0.990 and accuracy of 0.964.

Conclusions:

• Plasma lipidomic analysis combined with machine learning offers a promising approach for diagnosing the cause of death in suspected AIHD cases.
• This methodology can accurately determine the pathological stage of AIHD, aiding forensic investigations.
• The identified lipid biomarkers and developed models hold potential for improving the accuracy and efficiency of AIHD diagnosis in forensic settings.