AI Tools for Structural Biology
From AlphaFold's structure predictions to generative design with RFdiffusion — the AI tools reshaping structural biology and how to use them together.
Why structural biology first
No field has been reshaped by AI more visibly than this one. AlphaFold’s structure predictions, now used by more than three million researchers across 190 countries, turned a problem that once took years of crystallography into a database lookup. That shift is the clearest existence proof that AI can move a scientific field forward — which is why it anchors this guide.
Core tools
AlphaFold (AlphaFold DB / AlphaFold3)
What it does: Predicts protein structure from sequence. AlphaFold3 extends this to predict interactions between proteins, DNA, RNA, and small molecules.
Access: Free — AlphaFold Database for pre-computed structures; open-use models for custom predictions.
Best for: Anyone needing a structure prediction without running their own crystallography — the default starting point for nearly any structural question.
Note: This is a research model/database, not a SaaS product — there’s no “sign up” flow, just database access and, for custom runs, code from the associated repositories.
RFdiffusion / ProteinMPNN
What it does: Generative design models — rather than predicting the structure of an existing protein, these design new protein/peptide structures from scratch to meet a target function.
Access: Open source / research access.
Best for: Synthetic biology and peptide-drug researchers doing de novo design, not structure lookup. Pairs naturally with AlphaFold: design with RFdiffusion/ProteinMPNN, then validate structure with AlphaFold.
→ Full RFdiffusion / ProteinMPNN tool review
RoseTTAFold All-Atom
What it does: Generalized biomolecular modeling and design tool covering a broader set of molecular interactions than earlier single-purpose structure predictors.
Access: Open source / research access.
Best for: Researchers who need a broader modeling toolkit beyond protein-only structure prediction.
Isomorphic Labs’ drug-discovery engine (context, not a directory listing)
Isomorphic Labs — a DeepMind spin-off — has built a proprietary successor system for protein–drug interaction prediction, described by outside experts as a major advance on the scale of “an AlphaFold 4.” Unlike AlphaFold, this system is not open or independently accessible to outside researchers, so it isn’t listed as a usable tool here. It’s included for context: if you’re tracking where the field is headed, proprietary drug-discovery engines built on AlphaFold’s foundations are a trend worth watching, even though you can’t use this one directly yet.
Suggested workflow
| Step | Tool | Purpose |
|---|---|---|
| 1 | AlphaFold DB | Check if a structure prediction already exists for your target |
| 2 | RFdiffusion / ProteinMPNN | Design a novel protein/peptide if no suitable existing structure fits your need |
| 3 | AlphaFold3 | Validate structure and predict interactions for your designed candidate |
| 4 | RoseTTAFold All-Atom | Cross-check complex multi-molecule interactions if needed |
What these tools can’t do yet
- They predict structure, not function with certainty. A confident structure prediction doesn’t guarantee correct biological behavior — experimental validation is still required before strong claims.
- Generative design tools produce candidates, not guarantees. RFdiffusion/ProteinMPNN outputs need wet-lab testing; treat them as a hypothesis-generation step, not a final answer.
- Access and openness vary a lot within this field. AlphaFold is fully open; Isomorphic’s newer engine is proprietary. Don’t assume every “AlphaFold-adjacent” tool you hear about is equally accessible.
Related content
- Tool: AlphaFold
- Tool: RFdiffusion / ProteinMPNN
- Related field guide: Materials Science (coming soon)