Machine-learning-aided identification of steroid hormones based on the anisotropic galvanic replacement generated sensor array

Abstract

The conveniently simultaneous identification of steroid hormones is challenging due to their similar chemical structures. Herein, we have established a machine-learning-aided sensor array for accurate discrimination and determination of steroid hormones based on [email protected]2O (CC), [email protected]2[email protected] (CCP) and [email protected]2[email protected] (CCPA). The Michaelis-Menten constant of CC, CCP and CCPA toward H2O2 were calculated as 2.61, 1.11 and 2.03 mM, respectively, indicating their different catalytic activities that were obtained from the anisotropic galvanic replacement of [email protected]2O with Pd (II) and Au (III). Five kinds of steroid hormones were selected as model targets and reacted with the sensor array before chromogenic reaction. The absorption of chromogenic substrate was used as learning data to train the k-nearest neighbors algorithm, the discrimination confidence was from 88.9% to 100% for different mixtures, and 100% for betamethasone of 1–50 μM in real sample. This work provides a quick analysis of steroid hormones in 1.5 h and low-cost strategy, its further application in the field of cosmetic safety is highly expected.