AlphaFold 3
Predict the structure of proteins, DNA, RNA, and ligands — and how they interact.
About
AlphaFold 3 (released May 2024) predicts the 3D structure of proteins plus DNA, RNA, small-molecule ligands, ions, and modified residues — jointly, in one model. It uses a diffusion-based architecture with a module called Pairformer, and on protein-ligand interaction benchmarks it far outpaced earlier specialized tools.
Most researchers use it through the free AlphaFold Server, which needs no ML setup. Code and weights were released for academic (non-commercial) use in November 2024; commercial drug-discovery use runs through Isomorphic Labs. Practical limits: the server caps free use at around 30 jobs per day and offers a smaller ligand list than the full model.
What it can do
- Joint structure prediction for proteins, DNA, RNA, ligands, ions
- Protein-ligand interaction modeling
- MSA and template-based prediction
- Confidence scoring per residue
Setup
- Go to alphafoldserver.com and sign in with a Google account.
- Paste protein sequences and add DNA, RNA, or ligand inputs from the available lists.
- Submit the job and wait for the prediction.
- Download the predicted structures and confidence scores.
- For local runs: clone the GitHub repo, accept the terms of use, request model weights, and run the inference pipeline on a GPU machine.
Prerequisites
Google account for the server; GPU + large disk/RAM for local inference.
Use cases
Works with
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: ubiquitous in structural biology — standard tool in the field; actively maintained by Google DeepMind; server and code updated through 2026; code and weights available for academic use only; commercial use gated through Isomorphic Labs. Manual review: predictions are state of the art, but treat low-confidence regions and novel ligands with skepticism. License: Code under CC-BY-NC-SA 4.0 for academic use; commercial via Isomorphic Labs..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 10.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 9.0/10
- · ubiquitous in structural biology — standard tool in the field
- · actively maintained by Google DeepMind; server and code updated through 2026
- · code and weights available for academic use only; commercial use gated through Isomorphic Labs
- · manual review: predictions are state of the art, but treat low-confidence regions and novel ligands with skepticism
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.