register: watson-rfdiffusion-2023-gaia 0.1.0

[kunyuan]

Register: watson-rfdiffusion-2023-gaia 0.1.0

• Repository: https://github.com/kunyuan/watson-rfdiffusion-2023-gaia
• Tag: v0.1.0
• IR Hash: sha256:ac3252109c2e000554c322b9934c31d0096fc926aa80222d425f722f9780f585

Package summary

Formalization of Watson et al. 2023, "De novo design of protein structure and function with RFdiffusion" (Nature).

• 131 knowledge nodes (14 settings, ~117 claims)
• 43 strategies (19 noisy_and, 21 abduction, 3 induction)
• 8 modules: motivation, method, unconditional generation, symmetric oligomers, motif scaffolding, symmetric motif scaffolding, binder design, discussion
• Self-review: 52 leaf priors, no derived claim priors (inference engine defaults); beliefs determined entirely by BP

Exported claims (7)

Export	Belief	Summary
symmetric_high_success	1.000	Symmetric oligomers with any point group symmetry
rfdiffusion_benchmark_performance	1.000	23/25 motif scaffolding benchmark problems solved
binder_success_rate	1.000	19% binder success rate, ~100× over Rosetta
comprehensive_improvement	0.993	Comprehensive improvement over prior methods
ha20_atomic_accuracy	0.875	Atomic-level design accuracy (0.63 Å cryo-EM)
rfdiffusion_broad_success	0.763	Outstanding performance across all design tasks
generality_claim	0.611	Protein design from minimal specifications

🤖 Generated with Claude Code