Repository navigation
Home
Predict small-molecule bioactivity from structure, explore compounds, and build classifiers from labeled screening data.
Open MolClass — browse compounds and use published models without installing software.
Search · Published models · Dataset review · Molecule directory
Live compound-registry screenshot, captured 3 November 2026.
| Step | Guide | What it covers |
|---|---|---|
| 1 | Overview | Purpose, useful applications, and interpretation limits |
| 2 | Using MolClass | Search, molecule cards, selection, prediction, and new structures |
| 3 | Current datasets | All 21 hosted datasets, descriptions, labels, counts, and references |
| 4 | Build and review models | SDF import, features, classifiers, evaluation, and publication |
| 5 | REST API and MCP | Programmatic access and integration examples |
| 6 | Installation and operations | Docker setup, resources, approval setup, and maintenance |
| 7 | Architecture | Services, data flow, and versioned contracts |
| 8 | FAQ | Scores, dataset eligibility, progress, and common problems |
The interface is branded MolClass V2. The maintained registry, model lifecycle, and prediction API use v3 contracts. Management endpoints use /api/v1; this prefix does not mean they are the retired v1 application.
These pages were revised on 3 November 2026 against the current working codebase and read-only live API responses. Live dataset/model counts are dated snapshots. Source capabilities and deployed settings can differ; the documentation update notes identify the important differences.
The repository documentation index provides detailed engineering guides. Historical wiki revisions remain available in Git history.