ChemSpaceAL
A GPT writes molecules, a sample from each cluster is docked and scored, and the model is fine-tuned on the best clusters.
The loop
- Proposes
SMILES strings from a GPT-based molecular generator (100,000 unique molecules per iteration)
- Judged by
About 1,000 molecules sampled across clusters are docked with DiffDock and scored by contact-based interactions
- Steers
Clusters are weighted by mean score into a training set, and the GPT is fine-tuned on it
Headline result
In the paper's computational study on c-Abl kinase, the model learned to generate molecules similar to FDA-approved inhibitors without being shown them and reproduced two exactly (imatinib and bosutinib after five iterations). The method was also applied to the HNH domain of Cas9, which has no commercial small-molecule inhibitors.
What this does not prove
The judge is a learned docking model (DiffDock) plus a simple contact-count score, so this is predicted binding only. Matching known drugs is a retrospective sanity check, and no molecule was tested in a lab.
How it generalizes the harness
A SMILES GPT (not an agent LLM) is fine-tuned each round on docking-scored samples, so feedback changes the generator's weights; the judge is a learned docking model, not a physical authority.
Classification
| What it proposes | Molecule |
|---|---|
| Generative model | GPT-style transformer trained on SMILES strings, fine-tuned each iteration |
| Algorithmic authority | Docking (DiffDock) with an interaction-based scoring function; cluster-based sampling |
| Feedback | Score |
| What gets updated | Model weights and training data |
| When it runs | During training |
| Validity | Surrogate (a score or model stands in for the real test) |
| Evidence | Experiments reported |
| Lineage | Original system |