Integrated metabolomics and transcriptomics reveal biomarkers for detecting fentanyl analog abuse through machine learning.

Fentanyl analogs' (FA) rampant proliferation causes widespread abuse, fatalities, and severe social issues globally. Their high toxicity, rapid metabolism, and poor detectability make them a major anti-drug challenge. This study aimed to establish an accurate method for identifying their abuse. Firstly, a rat model of FA (fentanyl, remifentanil, and sufentanil) abuse was constructed using the conditioned place preference (CPP) paradigm. Subsequently, dual-platform (Liquid and Gas Chromatography - High Resolution Mass Spectrometry) based metabolomic and transcriptomic data were used for weighted gene co-expression network analysis (WGCNA) and trait-module correlation analysis, key genes and metabolites associated with overall FA abuse were identified. Thirdly, we established a FA abuse identification model (100 % accuracy, AUC=1) based on 31 metabolites and 12 hub genes. Finally, we established an accurate classification model of different FA (77.27 % accuracy; AUCs: 1, 1, 0.981, 0.979 for control/fentanyl/remifentanil/sufentanil groups) using ten genes and five metabolites. The identified biomarkers and computational models offer valuable tools for clinical diagnosis and drug regulation, which are critical for addressing the ongoing opioid crisis.