Abstract
<title>Abstract</title> <p>Objectives Antimicrobial resistance continues to increase the need for computational resources that support antibacterial drug discovery. This study compiled and curated a dataset of more than 220 β-lactam, azetidinone, and thiazolidinone derivatives collected from published literature reporting experimentally determined minimum inhibitory concentration (MIC) values. The objective was to produce a standardized dataset containing molecular structures, calculated molecular features, and antimicrobial activity classifications that can support quantitative structure–activity relationship (QSAR) studies and machine learning-based antimicrobial research. Data description The dataset includes standardized molecular structures represented using Simplified Molecular Input Line Entry System (SMILES) notation, experimentally determined minimum inhibitory concentration (MIC) values, multiclass antimicrobial activity labels, molecular fingerprints, and physicochemical descriptors generated using the RDKit cheminformatics toolkit. Data processing included structure standardization, descriptor calculation, feature integration, and activity classification. The dataset provides a reusable resource for developing, benchmarking, and validating machine learning models for antimicrobial activity prediction while supporting reproducible computational drug discovery research.</p>