Machine learning-assisted smart electrochemical platform: High-sensitivity simultaneous detection of rutin and...

Rutin (Rn) and luteolin (Lu), as important plant-derived natural compounds, hold significant value in the fields of medicine, nutrition, and food science. However, their highly similar structures often lead to challenges such as signal overlap and poor selectivity in traditional electrochemical detection methods. To address these issues, this study developed an intelligent electrochemical sensing platform that integrates nanocomposites and machine learning. A ternary heterojunction structure of biochar (CB)/ZIF-8/MnIn₂S₄ was employed to synergistically enhance electron transfer and catalytic efficiency. Combined with the random forest algorithm, the platform enabled machine learning-assisted feasibility assessment of target analytes, optimization of experimental parameters, and precise concentration prediction. The sensor demonstrated a wide linear range for Rn (0.01-500 μM) and Lu (0.29-784.8 μM), low detection limits (Rn: 3.48 nM, Lu: 2.41 nM), and high sensitivity (5.54 μA μM^-1 cm^-2). Furthermore, the random forest model achieved high-precision mapping of signal-concentration relationships, with a high classification accuracy. In real sample analysis (honeysuckle, ginkgo leaf extracts, and human serum), the recovery rates ranged from 96.4% to 105.2%, with an RSD of less than 3.72%, consistent with HPLC results. This work provides a reliable data-driven sensing approach for the quality control of natural product active ingredients and with potential for clinical monitoring of blood drug concentrations.