Overview
CYPMode combines molecular features and predictive models for time-dependent inhibition with a separate CYP3A4 structure workflow. The method report describes nested cross-validation, predefined selection gates and uncertainty reporting. Its negative-results ledger records approaches that failed those gates. These are author-reported computational methods; training-data estimates are not hidden-test scores or clinical validation.
The structural pipeline builds Boltz-2 inputs from ligand SMILES, invokes an external predictor and summarizes coordination geometry. This requires a separate Boltz environment and computational resources. Some challenge outputs and the full notebook are held back until the challenge closes. Code is AGPL-3.0-or-later with NOTICE attribution terms; documents are CC BY-NC-SA 4.0, generated data CC BY 4.0, and third-party material retains its own terms. The project is a challenge-related research contribution; no organizer endorsement is established.
Key Features
- TDI modeling for CYP3A4 and CYP2D6 with an auditable method report.
- A ledger of unsuccessful candidate methods alongside selection decisions.
- SMILES-to-Boltz-input structural pipeline and result summaries.
Use Cases
- Study documented evaluation choices for CYP inhibition models.
- Inspect negative results before extending a research pipeline.
- Explore CYP3A4 ligand coordination hypotheses using predicted structures.
How to Use
Use the current repository documentation and preserve its NOTICE and provenance records when reusing the work. The package metadata requires Python 3.10 or later. The quick start documents a virtual environment and pip install -e ".[challenge,validate]" from a source checkout. The app is launched with streamlit run app.py. For new structural predictions, configure the separate Boltz environment before using cypmode.validation.pipeline; its default accelerator is GPU. Public summaries can be inspected without assuming every challenge output is released. These commands are documentation examples only; no code or scientific calculation was run.
Sources and verification
This entry describes capabilities documented by the linked project sources. Catalog edits are not evidence of a successful run, compatibility test or scientific performance evaluation.
Read the editorial and verification policy