AI-SNPS2: A Multi-Layered LC-MS/MS Platform Integrating Molecular Networking and Retention Time Prediction for...

Area of Science:

• Forensic Science
• Analytical Chemistry
• Computational Chemistry

Background:

• Emerging unknown controlled substances present significant challenges for forensic and analytical sciences.
• Traditional liquid chromatography-tandem mass spectrometry (LC-MS/MS) is limited in identifying novel synthesized analogues not present in existing spectral libraries.

Purpose of the Study:

• To develop an enhanced screening software, AI-SNPS2 (Artificial Intelligence Screener for Narcotic Drugs and New Psychoactive Substances 2), to address the limitations in identifying unknown controlled substances.
• To extend identification capabilities beyond conventional spectral search by integrating structural analogue detection via spectral similarity and chromatographic plausibility filtering.

Main Methods:

• AI-SNPS2 integrates five layers: LC-MS Viewer, AI Classifier, Identifier, NetBuilder (molecular networking), and RT Predictor (retention time prediction using machine learning).
• The RT Predictor layer utilizes four regression models (ANN, SVR, RF, XGBoost) trained on molecular descriptors from 164 controlled substances.
• Software utility was evaluated by spiking known synthetic cannabinoids (JWH-019, JWH-015, JWH-302) into complex matrices and analyzing additional compounds.

Main Results:

• The RT Predictor models demonstrated strong performance, with XGBoost achieving the highest accuracy (R² = 0.964, MAE = 0.585).
• RT prediction deviations were within minutes on a 110 min gradient, proving its utility for candidate filtering.
• AI-SNPS2 successfully identified spiked synthetic cannabinoids using molecular networking and hybrid similarity search, and demonstrated promise in detecting compounds absent from databases.

Conclusions:

• AI-SNPS2 offers an advanced approach for identifying novel psychoactive substances and controlled analogues by combining spectral similarity, molecular networking, and retention time prediction.
• The software significantly enhances the capabilities of forensic and analytical laboratories in detecting and identifying unknown compounds, improving public safety.