Tutorial Map

COFOLDER tutorials come in two forms:

• written tutorials in docs/tutorials/ for the shortest supported reading path
• marimo notebook tutorials in the repository tutorials/ folder for interactive exploration

Suggested Learning Path

1. Copy the distributable inputs with cofolder-tools copy-examples ./cofolder-examples.
2. Read Quick Start.
3. Read User Guide Overview.
4. Work through Basic Usage.
5. Continue with the workflow-specific or topic-specific tutorials below.

Written Tutorials

• Basic Usage: first runner-backed workflow
• Virtual Screening: screening across a ligand library
• Ligand Handling: ligand formats, conformers, and utilities
• Advanced Features: advanced configuration and power-user features
• MAPK14 Structure-gated Oracle: pose-aware ranking by binding mode, designated interactions, and affinity score
• Adding New Runners: contributor-facing runner authoring guide
• Backend Acceptance: clean-install validation for backend and CLI changes

Notebook Tutorials

These live in the repository root under tutorials/:

• bias.py
• validate.py
• screen.py
• oracle.py
• structure_gated_oracle.py
• ligand_handling.py
• runners.py
• boltz_system_inputs.py
• openfold3_system_inputs.py

Launch them with marimo:

python -m pip install -e ".[analysis,tutorials]"
marimo edit tutorials/bias.py

Use the notebooks when you want to inspect intermediate outputs, execute cells step by step, or adapt the tutorial flow to your own systems. The workflow-first set centers on bias, validate, screen, and oracle, then adds cross-cutting ligand handling plus one contributor notebook for runners and backend acceptance.

Release Scope

The 1.0 tutorial surface covers standalone bias evaluation, Validate, fixed-system ligand-library Screen, Oracle, and SMILES/SDF/MOL/CCD ligand handling. Protein-iteration or selectivity screening, ligand soaking, and standalone molecule generation are post-v1 ambitions and are not presented as implemented workflows.

Real-Backend System Inputs

The two system-input notebooks introduce DNA, RNA, and top-level constraints. Real inference runs only after you enable a checkbox. A successful run displays the exact backend input and checks the normalized chain order; a separate safe example shows an actionable preflight failure without launching inference.

Use a dedicated CUDA-capable environment for the selected backend:

python -m pip install -e ".[analysis,tutorials,boltz2]"
marimo edit tutorials/boltz_system_inputs.py

The Boltz notebook also supports the boltz1 and boltz-community extras through its runner selector. Install only one Boltz package line in an environment.

OpenFold3 additionally requires its downloaded cache and setup marker:

python -m pip install -e ".[analysis,tutorials,openfold3]"
export OPENFOLD_CACHE="$PWD/cache/.openfold3-cache"
cofolder-tools setup-openfold3
marimo edit tutorials/openfold3_system_inputs.py

These notebooks teach the user workflow. Maintainers should still use Backend Acceptance to record release acceptance across every claimed backend route.