Calibration-free on-site detection of microcystin-LR using integrated biosensing, multi-parameter water quality monitoring, and machine learning.

Microcystin-lysine-arginine (MC-LR) is a potent hepatotoxic cyanotoxin produced during harmful algal blooms and is recognized as a global environmental concern due to its widespread occurrence in surface waters. Screen-printed carbon electrode (SPCE) biosensors offer a rapid and low-cost approach for MC-LR detection. However, their electrochemical responses are strongly influenced by water quality conditions (e.g., pH, electrical conductivity, total dissolved solids), often necessitating frequent, condition-specific recalibration. To address this limitation, this study developed an integrated framework that combines SPCE-based antibody biosensing with multi-parameter water quality monitoring and machine learning (ML) for rapid and on-site detection of MC-LR. Field samples collected across Florida yielded over 200 observations integrating biosensor impedance responses with concurrent measurements of pH, EC, turbidity, and UV254. Among the evaluated ML models, Extreme Gradient Boosting (XGBoost) achieved the highest predictive performance (NSE = 0.89, RMSE = 13.21), enabling a unified model applicable across diverse water matrices. Shapley Additive Explanations (SHAP) analysis identified impedance (Z') as the dominant predictive feature, followed by EC, pH, UV254, and turbidity, highlighting the influence of matrix-driven interfacial effects on biosensor response. Overall, this work establishes a robust, data-driven framework for characterizing biosensor-water matrix interactions and provides a practical strategy for faster and more accurate on-site measurement of MC-LR in complex environmental waters.