Prioritization of Unknown LC-HRMS Features Based on Predicted Toxicity Categories

Area of Science:

• Environmental chemistry
• Analytical chemistry
• Toxicology

Background:

• Liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS) enables nontargeted analysis (NTA) of complex environmental samples.
• Identifying individual constituents in NTA is challenging due to the large number of detected features.
• Prioritization strategies are crucial for focusing on relevant features in environmental sample analysis.

Purpose of the Study:

• To develop a novel prioritization strategy for environmental sample analysis that directly links fragmentation and chromatographic data to aquatic toxicity categories.
• To bypass the need for complete identification of individual compounds for toxicity assessment.
• To create robust models for predicting aquatic toxicity categories from LC-HRMS data.

Main Methods:

• Developed a Random Forest Classification (RFC) model using MS1, retention, and fragmentation data (cumulative neutral losses - CNLs) for toxicity prediction when fragmentation data is available.
• Developed a Kernel Density Estimation (KDE) model utilizing only retention time and MS1 data when fragmentation information is absent.
• Evaluated model performance on a pesticide mixture in a tea extract using LC-HRMS.

Main Results:

• Both RFC and KDE models demonstrated accuracy comparable to structure-based prediction methods.
• The CNL-based RFC model achieved an accuracy of 0.76 in real-world applications.
• The KDE model achieved an accuracy of 0.61, demonstrating its utility when fragmentation data is limited.

Conclusions:

• The developed prioritization strategies effectively link LC-HRMS data to aquatic toxicity categories without requiring compound identification.
• These models offer a robust and efficient approach for prioritizing features in environmental sample analysis.
• The models show strong performance in practical applications, enhancing the study of complex environmental matrices.