COFOLDER Documentation¶

COFOLDER provides command-line workflows for protein-ligand co-folding and pre-cofolding diagnostics. This site documents COFOLDER 1.0.0, released under the MIT licence from the CDDLeiden/COFOLDER repository.

The supported user-facing workflows are:

bias: standalone reference-overlap diagnostics from system.yaml
validate: single runner-backed co-folding workflow
screen: validate-style workflow across a ligand library
oracle: single-metric runner-backed scoring workflow

Start Here¶

If you are new to COFOLDER, follow this order:

Choose Your Path¶

I want the shortest path to a first run¶

Read Quick Start
Copy files from examples/
Run cofolder validate

I want to understand which command to use¶

Read User Guide Overview
Then jump to bias, validate, screen, or oracle

I want step-by-step learning¶

Start with Basic Usage
Continue through the Tutorial Map

I want contributor-facing backend context¶

Read Adding New Runners
Use Backend Acceptance before promoting backend or CLI-adjacent changes

I want to use the Python API¶

Follow the Bias recipe
Follow the Validate recipe
Follow the Screen recipe
Follow the Oracle recipe
Read Constructing a new backend runner

Repository Orientation¶

The public repo is organized around a few distinct user paths:

docs/: curated reading path
examples/: copy-and-run input files
repository tutorials/: interactive notebooks
scripts/: optional setup and bias-data helpers

For a concise map of those roles, see Repository Tour.

Scientific Scope¶

COFOLDER keeps validation metrics, model-derived confidence metrics, and structure-derived diagnostics distinct. Structural confidence is not a proxy for binding affinity.

Machine-readable citation metadata is available in the repository's CITATION.cff.