RoseTTAFold All-Atom
One network for proteins, nucleic acids, small molecules, metals, and covalent modifications, with design built in.
About
RoseTTAFold All-Atom (RFAA), published in Science in March 2024 by Krishna, Wang, and colleagues, generalizes protein structure prediction to full biomolecular assemblies. Built on the RoseTTAFold2 architecture, it represents protein and nucleic acid chains as sequences and small molecules, metals, and covalent modifications as atom-bond graphs, adding element types, bond orders, and chirality into the three tracks. An all-atom version of the frame-aligned point error loss trains it, and it predicts per-atom and pairwise confidence like AlphaFold.
The paper reports protein monomer accuracy comparable to AlphaFold2, strong results in the CAMEO blind ligand-docking challenge, and accurate modeling of covalent modifications. Fine-tuning the same network for diffusion produced RFdiffusion All-Atom, which designed and experimentally validated binders for digoxigenin (10 nM affinity), heme, and bilin molecules. Code and paper weights are on GitHub, but the weights carry non-commercial terms under the Rosetta-DL license family.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Author-reported (Krishna et al., Science 2024): wins over baseline pipelines in the CAMEO blind ligand-docking challenge; 42% of complexes predicted correctly vs other deep-learning docking methods; 46% of covalent modifications under 2.5 Å RMSD; protein monomer accuracy comparable to AlphaFold2; RFdiffusionAA binder designs experimentally validated
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (metrics tier, metrics-v1). Evidence: Science 2024 paper (DOI 10.1126/science.adl2528) and repo github.com/baker-laboratory/RoseTTAFold-All-Atom confirm the method, the CAMEO results, and the experimentally validated RFdiffusionAA designs. Strengths: first general all-atom biomolecular network, strong flexible-backbone docking results, real wet-lab validation of designed binders. Docked for: (1) weights restricted to non-commercial use under Rosetta-DL terms; (2) heavy compute and a complex MSA/template database setup; (3) still new, with limited independent benchmarking so far. License: Rosetta-DL family (non-commercial weights); code available on GitHub.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 6.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · First general all-atom biomolecular network; Science 2024 with experimentally validated RFdiffusionAA designs.
- · Weights restricted to non-commercial use.
- · Heavy compute and complex MSA/template database setup; limited independent benchmarking so far.