LASSBio-classFLOW: A Semiautomated KNIME Workflow for Classification Model Benchmarking and Virtual Screening

Abstract Machine learning-based quantitative structure–activity relationship (ML-QSAR) modeling requires consistent data preparation, validation, and model comparison. We present LASSBio-classFLOW, an open, modular, and semiautomated KNIME workflow for classification-based ligand screening. It integrates basic molecular structure preparation, configurable class assignment, descriptors and fingerprints calculation, dataset partitioning, hyperparameter optimization of nine algorithms, interactive performance evaluation, user-guided model selection, and external-library prediction. As an example, the workflow was evaluated using 1408 ROCK2 compounds from ChEMBL (218 active and 1190 inactive). Models were trained and optimized by fivefold cross-validation and evaluated on a held-out 20% test set. Support vector machine (SVM) and k-nearest neighbors (kNN) provided the most consistent results across both stages. On the held-out set, SVM favored active-class precision (0.861), whereas kNN achieved higher active recall (0.750). LASSBio-classFLOW (available at https://github.com/pedrosenamp/LASSBio-classFLOW_v1.0.git) therefore provides a transparent, reusable environment for benchmarking, comparative ML-QSAR development, and flexible deployment, while preserving user control over model selection and screening decisions.