An emerging machine learning-optimized MIP-ECL platform for ultrasensitive and selective detection of 5-hydroxytryptamine.

Accurate detection of 5-hydroxytryptamine (5-HT) is essential for the diagnosis and treatment of clinical depression. A functionalized film was fabricated by electropolymerizing graphene oxide (GO) and l-arginine (L-Arg) hybrid onto the electrode surface, followed by gold nanoparticles (AuNPs) deposition to form a AuNPs/Poly-(L-Arg)/rGO nanocomposite. This nanocomposite served as the substrate for a molecularly imprinted polymer (MIP)-based electrochemiluminescence (ECL) sensing interface for the ultrasensitive and selective detection of 5-HT. Considering the obvious significance of MIP fabrication parameters on 5-HT detection performance, an orthogonal experimental design (OED) was initially employed to evaluate the effects of four critical parameters. Subsequently, an XGBoost regression model was applied to establish the nonlinear relationship between fabrication parameters and occupancy rate of imprinted cavities. Bayesian optimization (BO) was applied to globally optimize the multidimensional parameter space with the objective of maximizing cavity occupancy to develop a performance-optimized MIP sensor for 5-HT detection. Under optimized conditions, linear detection of 5-HT over a concentration range of 0.01-100 μM was accomplished using a cathodic Ru(bpy)32+/K2S2O8- based ECL system operating via a redox mechanism, with detection limits down to 7.04 nM. A significant inverse correlation was observed between ECL intensity and the logarithm of 5-HT concentration (R2 = 0.99). Moreover, the sensor demonstrated excellent anti-interference capability, repeatability, and reproducibility, and was successfully applied to the accurate detection of 5-HT in spiked fetal bovine serum samples.