Multi-Modal Foundation · recursionpharma
Nesso
Predicts protein–ligand binding affinities and binder status from protein sequences and small molecules.
What Nesso is
Nesso, also known as Nesso-1, is a structure-based drug discovery model developed by Valence Labs at Recursion. It evaluates interactions between proteins and small molecules directly from sequence and chemical structural representations.
The model addresses computational bottlenecks in structure-based drug design by applying a coarse-grained co-folding approach. This co-folding strategy accelerates the prediction of protein–ligand binding affinity relative to conventional structural methods.
Given a target protein sequence and a candidate ligand, the system outputs both a continuous binding affinity scalar and a discrete binder/non-binder classification score.
How it works
Nesso uses a coarse-grained co-folding framework to predict binding characteristics directly from protein amino-acid sequences and small-molecule chemical descriptions.
Key facts
- Model type
- Coarse-grained co-folding model
- Input modalities
- Protein amino-acid sequence and ligand (SMILES, CCD code, or SDF)
- Outputs
- Binding affinity scalar and binder/non-binder score
- Category
- Drug discovery
What you can use it for
- Binding affinity estimation
- Estimating numerical binding affinity scalars for protein–ligand pairs using protein amino-acid sequences and ligand definitions.
- Virtual screening triage
- Classifying compound libraries into binders and non-binders to filter molecules for downstream evaluation.
- Structure-based drug design
- Accelerating computational exploration of chemical series against specific protein targets.
Run it here
We list this model but cannot run it yet: unsupported task: other.
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- Combines continuous affinity prediction with binary binder/non-binder classification in a single model.
- Accepts multiple common ligand input formats, including SMILES, Chemical Component Dictionary (CCD) codes, and SDF files.
- Accelerates binding-affinity calculation by employing coarse-grained co-folding.
Watch out for
- Restricted to inputs provided as linear protein amino-acid sequences and specific ligand formats (SMILES, CCD, or SDF).
- Requires structured input configurations such as prediction YAML schemas for CLI execution.
Inputs and outputs
You provide
Sequence or text
A protein or nucleotide sequence, or free text — whichever this model was trained on.
You get back
Mean-pooled representation, one vector per input.
Licence
Check the licence before commercial use · apache-2.0
The apache-2.0 licence carries terms worth reading before you depend on it.
Questions
Can I use Nesso commercially?
Check the licence before commercial use. The apache-2.0 licence carries terms worth reading before you depend on it.
What input data does Nesso require?
The model requires an amino-acid sequence for the protein target and a structural representation of the ligand provided as a SMILES string, a CCD code, or an SDF file.
What outputs are produced by a prediction run?
Nesso returns a predicted binding affinity scalar along with a binder/non-binder classification score.
How does Nesso differ from standard molecular docking tools?
Nesso performs coarse-grained co-folding directly from the protein sequence and ligand structure to accelerate affinity prediction, rather than relying on full-atom docking against pre-existing static structures.
How to cite it
Valence Labs. Nesso-1 Technical Report. 2026. https://www.valencelabs.com/wp-content/uploads/2026/07/nesso1.pdf
Written from Nesso's own documentation (model card). Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.