RANQSAR ML Suite - A desktop QSAR modeling software with ML, validation, AD, and reporting

RANQSAR v1.01 is a portable, end-to-end machine learning–driven Quantitative Structure–Activity Relationship (QSAR) platform designed for rapid, reproducible, and user-friendly predictive modeling of small-molecule bioactivity. The software integrates automated data curation, robust molecular featurization, leakage-safe model training, applicability domain assessment, and uncertainty estimation within a unified graphical interface. RANQSAR aims to bridge the gap between advanced QSAR methodology and practical usability by providing a fully self-contained environment that requires no external cheminformatics software. The platform supports Morgan fingerprint, curated 2D descriptor, and hybrid feature representations, enabling flexible modeling strategies suitable for diverse drug discovery datasets. ✨ Key Features Portable architecture: Fully self-contained Windows executable with embedded Python runtime; no external installation required. Automated data quality control: Built-in SMILES validation, duplicate removal, and dataset cleaning pipeline. Multiple molecular representations: Morgan fingerprints RDKit 2D descriptors with automated curation Hybrid (fingerprint + descriptor) features Leakage-safe preprocessing: Training pipelines designed to prevent information leakage during cross-validation and testing. Machine learning models: Random Forest–based QSAR workflow with reproducible training and evaluation. Applicability Domain (AD) analysis: k-nearest neighbor similarity–based AD (fingerprint-safe) Leverage (Williams plot) AD for descriptor space Prediction uncertainty estimation: Tree-ensemble variance–based confidence assessment. Reliability labeling: Integrated High/Medium/Low reliability classification combining AD and uncertainty. Automated reporting: Generation of prediction tables, model metrics, and reproducibility artifacts. User-friendly GUI: Streamlit-based interface for rapid model development and virtual screening. 🧪 Typical Workflow Import dataset (SMILES + activity). Automatic QC and cleaning. Select feature type (Morgan / 2D / Hybrid). Train QSAR model with scaffold or random split. Evaluate model performance and applicability domain. Predict external compounds with uncertainty and reliability assessment. 🎯 Intended Applications RANQSAR is designed for: QSAR model development Virtual screening prioritization Lead optimization support Academic and industrial cheminformatics workflows Rapid prototyping of ML-based activity prediction 🖥️ System Requirements Windows operating system No external dependencies required (portable runtime included) 📌 Version Notes (v1.01) This release includes improvements in: leakage-safe preprocessing pipeline descriptor curation workflow fingerprint-appropriate applicability domain assessment prediction reliability framework executable stability and portability