REINVENT 4
with DockStream docking
A molecule-writing neural network is rewarded by a scoring function, which can include docking, and its weights are updated toward high scorers.
The loop
- Proposes
SMILES strings for new small molecules, sampled from an RNN or transformer generator
- Judged by
A user-defined scoring function; here a DockStream docking score (Glide) plus drug-likeness (QED)
- Steers
Reinforcement learning, run as staged (curriculum) learning, raises the probability of high-scoring molecules in the generator
Headline result
In the paper's demonstration on PDK1 (PDB 2XCH), 50 epochs of staged learning with batch size 128 produced 119 hits out of 6,400 generated molecules (1.9%), where a hit means a docking score of -8 kcal/mol or better and QED of 0.7 or higher; the hits spanned 103 generic scaffolds.
What this does not prove
The PDK1 run is described by the authors as a hypothetical virtual screening exercise, and a good docking score is a computed estimate, not measured activity. Docking is one optional scoring component; DockStream alone is only a docking wrapper, not an AlgoGen, and its repository is no longer maintained.
How it generalizes the harness
- The generator is not an LLM
- Feedback retrains the generator
- A domain authority judges
- The output outlives the model
The generator is a SMILES RNN or transformer rather than an LLM agent, docking acts as the judge, and the reward updates the generator's weights.
Classification
| What it proposes | Molecule |
|---|---|
| Generative model | RNN and transformer SMILES generators (priors pretrained, then fine-tuned by transfer and reinforcement learning) |
| Algorithmic authority | Scoring function with docking (Vina, rDock, Glide, GOLD via DockStream) and RDKit property checks |
| Feedback | Reward signal |
| What gets updated | Model weights |
| When it runs | During training |
| Validity | Approximate physics (simulation, not lab results) |
| Evidence | Code released |
| Lineage | Original system |