Related Experiment Video
Updated: Aug 11, 2026
06:54
Enhanced Extraction of Low-Molecular Weight DNA from Wastewater for Comprehensive Assessment of Antimicrobial Resistance
Published on: July 19, 2024
High-risk contaminants detected in wastewater effluent samples can be prioritized prior to structural assignment
Helen Sepman1, Lisa Jonsson2, Malte Posselt3
1Department of Chemistry, Stockholm University, Svante Arrhenius väg 16, Stockholm, 106 91, Sweden; Department of Environmental Science, Stockholm University, Svante Arrhenius väg 8, Stockholm, 106 91, Sweden.
Water Research
|August 4, 2026
Summary
This study introduces a new workflow using machine learning to prioritize hazardous pollutants from mass spectrometry data. This method speeds up risk assessment by predicting toxicity and ranking chemicals before complex identification.
Area of Science:
- Environmental Chemistry
- Computational Toxicology
- Analytical Chemistry
Background:
- Accurate identification of hazardous pollutants is crucial for environmental risk assessment but is often hindered by complex structural elucidation.
- Current methods for pollutant discovery are time-consuming and may lack accuracy in identifying toxic substances.
Purpose of the Study:
- To develop and validate a workflow for prioritizing detected chemical features from mass spectrometry data based on their potential environmental risk.
- To integrate machine learning models for predicting toxicity and ionizability, enabling risk-based ranking without additional standards.
Main Methods:
- A workflow was developed integrating MS2Tox and MS2Quant machine learning models to predict toxicity and ionizability from mass spectrometry data.
- A priority score (environmental concentration/lethal concentration) was used to rank features, serving as a proxy for risk.
- The workflow was validated using known chemicals and applied to wastewater effluent data.
Main Results:
- The workflow successfully classified chemicals into high and low risk categories with varying recall, precision, and accuracy depending on acquisition mode and species.
- Application to wastewater effluent identified 13-19% of features as "precautionary risk features" with a risk quotient ≥1.
- Spectral library matching identified 12% of prioritized features, while in silico tools showed discrepancies in high-risk feature structural annotation.
Conclusions:
- The developed data-driven workflow effectively prioritizes potentially hazardous pollutants from mass spectrometry data, streamlining the risk assessment process.
- The study highlights the utility of predictive toxicology and prioritization before structural elucidation, though structural assignment remains a bottleneck.
- This approach allows for efficient screening of complex environmental samples, focusing subsequent identification efforts on features with the highest potential risk.