Skip to content

Docs rfd3 adv enzyme tutorial - #379

Draft
rclune wants to merge 13 commits into
productionfrom
docs_rfd3_adv_enzyme_tutorial
Draft

rclune wants to merge 13 commits into
productionfrom
docs_rfd3_adv_enzyme_tutorial

Conversation

@rclune

@rclune rclune commented Sep 4, 2026

Copy link
Copy Markdown
Member

Addition of an advanced enzyme design tutorial that uses a metalloprotease as an example. Original tutorial draft by Seth Woodbury and Maruf Hasan, created for the 2026 Rosetta Commons Megathon event.

Some editing and formatting was done using Claude-Code.

rclune added 12 commits June 29, 2026 09:06
Created for the 2026 Tutorial Hackathon, authors are Seth Woodbury and Maruf Hasan.
Original document authored by Seth Woodbury and Maruf Hasan for the 2026 Tutorial Hackathon. Rachel Clune is editing for structure and length.
Add a new reference page (models/rfd3/docs/output.md) documenting all JSON output metrics produced by RFdiffusion3, including RMSD/MAE metrics, clash counts, secondary-structure fractions, H-bond fields, center-of-mass data, and conditional/partial-diffusion fields. Update the advanced enzyme design tutorial to include a detailed "Running RFdiffusion3" section with example CLI commands, recommended runtime flags (batching, guidance, jitter, seed, etc.), notes about generated trajectory files, and an "Analyzing the Outputs" section describing the output file types and how to inspect metrics and structures. These changes aim to help users run RFD3 reliably and interpret the per-design JSON metrics and trajectory outputs.
Rework the advanced enzyme design tutorial to improve clarity and add practical guidance. Split the outputs walkthrough into distinct sections (Final structure, Output JSON, Specification and Inference Sampler), add a PyMOL alignment tip, enumerate key output metrics to inspect, and include a Python filtering script example with recommended post-filter visual checks. Minor reorganizations and TODO placeholders remain in the file.
Introduce a new documentation page models/rfd3/docs/common_issues.md covering ligand code conflicts with the PDB Chemical Component Dictionary (CCD), including symptoms (Atom not found warnings and shape mismatch) and the recommended workaround (rename three-letter ligand codes to avoid CCD matching). Update models/rfd3/docs/index.rst to include the new page and remove a duplicated "Common Issues" heading from the advanced enzyme design tutorial to centralize troubleshooting guidance.
Add the output metrics guide to the docs index and expand it with the JSON file structure, metric conditions, and gotchas. Also add an advanced enzyme design tutorial section covering hydrogen bond conditioning.

The creation of output.md was aided by Claude-code. The file has been thoroughly reviewed and edited prior to this commit.
Enhances the advanced enzyme design tutorial with clearer internal cross-links to key terms, adds a new RASA conditioning section with parameter guidance and examples, and fills in/expands glossary definitions (including theozyme, ORI token, catalytic residue, and related terms). Also tightens a few wording and references updates for readability.
Added 'enzyme_design_tips' page and reorganized the RFD3 landing page to have a section for that information

Deleted emacs artifact files and and a CIF file that ended up not being used for this tutorial.
Clarify the default ligand chain validation behavior and the new allow_ligand_on_existing_chain option in the RFD3 input docs. This includes documenting the chain-collision check, the requirement for unique ligand chains, and per-chain ligand res_id reset behavior.

Also update the advanced enzyme design tutorial to enable the option, include the missing ligand atom set, and add guidance around relevant output metrics and trajectory files.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant