Multi-Modal Foundation · omtx
LULA-1.1
A sequence-only model predicting protein-ligand binding scores from amino-acid sequences and SMILES strings.
What LULA-1.1 is
LULA-1.1 is a lightweight machine learning model developed by Om Therapeutics (omtx) for drug discovery and biological research. It predicts binding scores between target proteins and small molecules using sequence-level representations alone.
By bypassing the requirement for three-dimensional protein structures, molecular docking simulations, or preliminary protein folding, the model simplifies the initial computational screening of small-molecule interactions against target protein sequences.
How it works
The model accepts an amino-acid sequence representing a target protein alongside a simplified molecular-input line-entry system (SMILES) string representing a chemical ligand. Without calculating or requiring three-dimensional spatial coordinates, docking poses, or folded tertiary structures, it evaluates the sequence pairing to produce a predicted binding score.
Key facts
- Domain
- Drug discovery, biology, and genomics
- Input formats
- Protein amino-acid sequence and ligand SMILES string
- Output
- Protein-ligand binding score
- Modelling approach
- Sequence-only (no structural inputs or docking)
What you can use it for
- Protein-ligand binding scoring
- Estimate the interaction affinity between a candidate small molecule and a target protein using only sequence strings.
- Small-molecule discovery screening
- Evaluate sets of SMILES-encoded compounds against target amino-acid sequences without performing computationally demanding docking routines.
- Unstructured or unfitted target assessment
- Score potential drug interactions for protein targets lacking resolved crystal structures or predicted three-dimensional conformations.
Run it here
We list this model but cannot run it yet: gated on the Hub — never run past a gate the user did not accept.
Everything on this page still applies. Running it needs a loader for this repository, not a bigger machine.
Strengths and limits
Good at
- Operates strictly on sequence data without requiring 3D structural coordinates, docking, or protein folding.
- Lightweight sequence-only design allows rapid binding scoring directly from primary sequence and SMILES inputs.
Watch out for
- Does not predict or model three-dimensional binding poses, contacts, or atomic coordinates.
- Cannot incorporate explicit structural data or conformational changes occurring upon ligand binding.
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 · other
This model is gated by its authors — its licence has to be accepted on the source repository before it will run.
Questions
Can I use LULA-1.1 commercially?
Check the licence before commercial use. This model is gated by its authors — its licence has to be accepted on the source repository before it will run.
What inputs are required to run LULA-1.1?
The model requires two inputs: the amino-acid sequence of the target protein and a SMILES string specifying the chemical structure of the ligand.
Does this model require 3D protein structures or docking software?
No. The model is sequence-only and does not use structural inputs, tertiary folding steps, or docking simulations.
What output does the model provide?
The model generates a numerical binding score for the supplied protein-ligand pair.
Written from LULA-1.1'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.