Machine Learning-based Classification for the Prioritization of Potentially Hazardous Chemicals with Structural...

Area of Science:

• Environmental Chemistry
• Analytical Chemistry
• Computational Chemistry

Background:

• Nontarget screening (NTS) using liquid chromatography-high-resolution mass spectrometry (LC-HRMS) is vital for identifying unknown environmental micropollutants.
• Prioritizing relevant features from complex LC-HRMS data is a significant challenge in NTS.
• Identifying potentially hazardous chemicals requires effective strategies for feature selection.

Purpose of the Study:

• To develop and evaluate a novel strategy for prioritizing NTS features based on structural alerts for potentially hazardous chemicals.
• To leverage tandem mass spectra (MS2) and machine learning models to predict the likelihood of features corresponding to chemicals with structural alerts.
• To assess the feasibility of this approach for aromatic amines and organophosphorus compounds.

Main Methods:

• Utilized raw tandem mass spectra (MS2) and machine learning models (neural network, random forest) for feature prioritization.
• Trained models on experimental MS2 data, focusing on fragments and neutral losses.
• Applied the developed models to prioritize LC-HRMS features in environmental surface water samples.

Main Results:

• A neural network model for organophosphorus structural alerts achieved an Area Under the Curve of the Receiver Operating Characteristics (AUC-ROC) of 0.97 and a true positive rate of 0.65.
• A random forest model for aromatic amines achieved an AUC-ROC of 0.82 and a true positive rate of 0.58.
• The strategy successfully prioritized LC-HRMS features in real-world surface water samples.

Conclusions:

• The developed strategy effectively prioritizes NTS features corresponding to potentially hazardous chemicals.
• Machine learning analysis of MS2 data offers a promising approach for enhancing chemical risk assessment in environmental monitoring.
• This method has high potential for further development and implementation in routine environmental analysis.