Machine Learning–Assisted Novel Electrochemical Biosensor Utilizing NbC@NC-ZIF-8/ADH Sensing Probe for Selective Detection of NADH in Wolfberry

Wolfberry, a medicinal and edible plant rich in nicotinamide adenine dinucleotide (NADH), is noted for its antiaging and antioxidant properties. However, the trace levels and instability of NADH present significant detection challenges. To overcome the inherent limitations of traditional electrochemical linear peak fitting, three machine learning algorithms were introduced: random forest, support vector machine regression (SVR), and long short-term memory neural networks. This pioneering approach established a complex nonlinear relationship model between concentration and response current, effectively mitigating high-concentration linear distortion and interference constraints within the detection system. The Shapley additive prediction method (SHAP) was ultimately employed to quantify the contribution of each electrode potential to the prediction of response currents. This study developed an electrochemical biosensor based on a NbC@NC-ZIF-8/ADH heterojunction substrate, combined with a machine learning–based nonlinear modeling approach, achieving precise quantitative detection of NADH concentration in wolfberry. It exhibits a broad detection range from 1 μM to 1.5 mM, with a detection limit as low as 0.075 μM. The linear prediction model achieved R 2  = 0.976. After constructing a nonlinear model using the SVR machine learning algorithm, the NADH concentration detection accuracy improved to R 2  = 0.996, significantly enhancing detection precision. SHAP analysis indicated that the 0.264 V potential contributed most significantly to the model output. This study established a trace detection research framework encompassing “heterojunction material preparation–density functional theory analysis–nonlinear model construction–SHAP interpretability analysis,” providing a reliable solution for the rapid and accurate detection of NADH in wolfberry. It demonstrates broad application prospects in the field of bioanalysis.