Convenient, precise, and high-throughput discrimination of multiple bioanalytes is of great significance for an early diagnosis of diseases. Array-based pattern recognition has proven to be a powerful tool to detect diverse analytes, but developing sensing elements featuring favorable surface diversity still remains a challenge. In this work, we presented a simple and facile method to prepare programmable metal-nanoparticle (NP)-supported nanozymes (MNNs) as artificial receptors for the accurate identification of multiple proteins and oral bacteria. The in situ reduction of metal NPs on hierarchical MoS<sub>2</sub> on polypyrrole (PPy), which generated differential nonspecific interactions with bioanalytes, was envisaged as the encoder to break through the limited supply of the receptor's quantity. As a proof of concept, three metal NPs, i.e., Au, Ag, and Pd NPs, were taken as examples to deposit on PPy@MoS<sub>2</sub> as colorimetric probes to construct a cross-reactive sensor array. Based on the principal component analysis (PCA), the proposed MNN sensor array could well discriminate 11 proteins with unique fingerprint-like patterns at a concentration of 250 nM and was sufficiently sensitive to determine individual proteins with a detection limit down to the nanomolar level. Remarkably, two highly similar hemoglobins from different species (hemoglobin and bovine hemoglobin) have been precisely identified. Additionally, five oral bacteria were also well separated from each other without cross-classification at the level of 10<sup>7</sup> CFU mL<sup>-1</sup>. Furthermore, the sensor array allowed effective discrimination of complex protein mixtures either at different molar ratios or with minor varying components. Most importantly, the blind samples, proteins in human serums, proteins in simulated body fluid environment, the heat-denatured proteins, and even clinical cancer samples all could be well distinguished by the sensor array, demonstrating the real-world applications in clinical diagnosis.