Protein Structure · DeepMind (served as BioNeMo NIM)
AlphaFold2
Predicts all-atom 3D protein structures from sequence data using multiple sequence alignments.
What AlphaFold2 is
AlphaFold2 is a protein structure prediction model developed by DeepMind. It generates three-dimensional structural models from primary amino acid sequences, acting as a reference standard for monomeric protein folding.
The system relies on deep multiple sequence alignments to achieve high structural accuracy. Because alignment generation is integrated into the workflow, predictions typically take minutes rather than seconds.
How it works
The model takes a protein sequence, incorporates multiple sequence alignment information, and computes all-atom three-dimensional coordinates alongside residue-level confidence estimates.
Key facts
- Version
- 2.3.2
- Domain
- protein-structure
- Task
- structure_prediction
- Input modality
- protein_sequence
- CASP14 benchmark
- 92.4 GDT_TS
- GPU memory
- 40 GB
What you can use it for
- Monomeric structure prediction
- Predicts all-atom three-dimensional coordinates in PDB format from a single protein sequence.
- Local model confidence scoring
- Provides per-residue pLDDT scores between 0 and 100 to evaluate which parts of the modelled backbone are reliably placed.
Run it here
No installation and no GPU of your own: fill in the inputs and it runs on our servers. Results stay in your browser unless you sign in.
Runs on Managed GPU endpoint
Runs on managed compute. Availability and startup time depend on current capacity.Run this model
AlphaFold2
Strengths and limits
Good at
- Serves as the accuracy benchmark for monomeric protein folding when deep sequence alignments are accessible.
- Demonstrated high accuracy with a CASP14 GDT_TS score of 92.4.
- Provides per-residue confidence estimates to directly indicate model reliability.
Watch out for
- Execution requires minutes rather than seconds due to the alignment search.
- Depends on the availability of a deep multiple sequence alignment for optimal folding accuracy.
- Requires 40 GB of GPU memory.
Inputs and outputs
You provide
Protein sequence
Single-letter amino acid code. Whitespace and FASTA headers are stripped automatically.
You get back
All-atom coordinates in PDB format, viewable and downloadable.
pLDDT score per residue (0–100). Above 70 is generally reliable.
Licence
Check the licence before commercial use · Apache 2.0 (code)
The Apache 2.0 (code) licence carries terms worth reading before you depend on it.
Questions
Can I use AlphaFold2 commercially?
Check the licence before commercial use. The Apache 2.0 (code) licence carries terms worth reading before you depend on it.
How do I run AlphaFold2 without installing anything?
Use the form under “Run it here” on this page: fill in the inputs and press Run model. It runs on our servers on 1× A100 40GB, typically in 3.0min, and a run costs about 490 credits. Visitors without an account get 1,000 credits a month.
What inputs does AlphaFold2 require?
The model requires an input protein sequence.
What outputs does the model produce?
It outputs all-atom structural coordinates in PDB format along with a per-residue pLDDT confidence vector.
How are the confidence scores interpreted?
The pLDDT score ranges from 0 to 100 for each residue, where values above 70 are generally considered reliable.
How long does a prediction take?
Inference takes minutes rather than seconds because of the multiple sequence alignment calculation.
Written from AlphaFold2's own documentation. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.