Machine Learning-Enhanced MEC Sensors with Feature Engineering for Quantitative Analysis of Multi-Component Toxicants

Area of Science:

• Environmental Science
• Analytical Chemistry
• Biotechnology

Background:

• Industrialization leads to complex mixed toxicant pollution.
• Conventional detection methods struggle with synergistic/antagonistic toxicant interactions.
• Existing methods are inadequate for analyzing multi-component toxicant mixtures.

Purpose of the Study:

• To develop an integrated framework using microbial electrochemical systems (MECs) and machine learning (ML).
• To quantify formaldehyde, tetracycline, Ag+, and Cu2+ in complex mixtures.
• To overcome limitations of traditional MECs in mixed toxicant analysis.

Main Methods:

• Integrated MECs with ML for toxicant detection.
• Generated dynamic current-time (I-t) signals from MECs.
• Employed mechanism-driven feature engineering to extract 22 multidimensional features.
• Utilized Random Forest (RF) model for quantification.

Main Results:

• RF model achieved R2 > 0.9 for all tested toxicants (formaldehyde: 0.959, tetracycline: 0.934, Ag+: 0.936, Cu2+: 0.957).
• Minimized Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE).
• Identified key microbial taxa, including Geobacter anodireducens and Comamonas testosteroni.

Conclusions:

• The developed framework overcomes traditional MEC limitations for mixed toxicant monitoring.
• Innovative feature engineering and ML integration enable accurate quantification.
• Provides a rapid, low-cost, high-accuracy tool for environmental mixed toxicant analysis.