Matrix: cover protein+DNA+RNA together, and nucleic-acid-only inputs - #118
Merged
Conversation
The UI offers protein, DNA and RNA inputs simultaneously and all four
diffusion tools support that combination, but no case supplied all three
together -- the most we covered was two entity types (protein+DNA,
protein+RNA, protein+DNA+ligand). That is the CRISPR-Cas9 shape
(protein + guide RNA + target DNA) and the configuration these
AF3-class models exist to predict.
- B11/O07/C07: protein + DNA + RNA on Boltz, OpenFold, Chai
- B12/B13: DNA-only and RNA-only on Boltz, pinning that protein is NOT
required. Verified as a supported category rather than an oversight:
AlphaFold 3 benchmarks 'low homology RNA-only or DNA-only complexes'
against RoseTTAFold2NA, and Boltz's parser (schema.py:1021) accepts
{protein,dna,rna,ligand} with no protein requirement.
All five pass in production. The confidence numbers are the interesting
part: nucleic-acid-only gives high pLDDT with near-zero pTM (B12 87.7/
0.13, B13 84.9/0.18) -- correct, since a lone 20-mer has local geometry
but no global fold to be confident about -- while the three-way complex
inverts it (B11 73.2/0.89).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DEWo4xvL59PHC1QUvtTV9F
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The UI offers protein, DNA and RNA inputs simultaneously, and all four diffusion tools (Boltz, OpenFold 3, Chai-1, ESMFold2) declare support for that combination — but no matrix case supplied all three together. The most we covered was two entity types at once.
That combination is the CRISPR-Cas9 shape (protein + guide RNA + target DNA) and is the configuration AF3-class models exist to predict.
New cases
Nucleic-acid-only is a supported category, not an oversight in our validation. Verified before adding the cases:
schema.py:1021) accepts{protein, dna, rna, ligand}and has no "at least one protein" check; the upstream docs describe each entity type independently.All five pass in production (
folding_260825.2). The confidence numbers are worth recording:The near-zero pTM on nucleic-acid-only is correct, not a failure: pTM measures global fold confidence, and a lone 20-mer with no partner has no global topology to be confident about, while its local backbone geometry is well determined (high pLDDT). B11 inverts exactly as theory predicts.
612 unit tests pass.