Designing molecules that score high on the DRD2 oracle

Dopamine receptor D2 is the target of most antipsychotics, and a support vector machine trained on measured activities is a standard stand-in for the assay. This task asks a design algorithm to find molecules that score highly on that model while calling it at most 10,000 times.

Task

Propose molecules as SMILES strings and maximise the DRD2 oracle, a support vector machine with a Gaussian kernel over ECFP6 fingerprints, fitted to ExCAPE-DB activity data and returning a score in $[0, 1]$. The run is scored by AUC top-10 over a 10,000-call budget, so both the final quality and how quickly it was reached count.

Background

The oracle was introduced with REINVENT in 2017 and became one of the most-used targets in generative chemistry. PMO put it in a budgeted setting after observing that most reported comparisons ignore how many oracle calls a method spends, which is the one resource that matters when the oracle is an assay rather than a model. DRD2 is the easy end of the benchmark: several methods already reach above 0.9, so the interesting question is how few calls suffice.

Data and evaluation

• Data: there is no training set to fit. Any database a method needs is restricted to ZINC 250k, which ships with the repository, so no method gets an advantage from a richer corpus.
• Budget and seeds: 10,000 oracle calls per run, and five independent runs with different random seeds (python run.py METHOD --task production --n_runs 5 --oracles drd2). Report the mean and standard deviation over the runs. The budget counts distinct valid canonical SMILES: a molecule queried twice is served from the cache for free, and an invalid SMILES scores 0.
• Metric: AUC top-10. The mean score of the ten best molecules found so far is recorded every 100 calls, the curve against the number of calls is integrated by the trapezoid rule and divided by the 10,000-call budget, so the value lands in $[0, 1]$ and higher is better. A run that stops early keeps its final top-10 value for the rest of the budget. Unlike a plain top-10 average, the metric rewards reaching a good score early.
• Official evaluation code: https://github.com/wenhao-gao/mol_opt (MIT), where main/optimizer.py holds the budget accounting and the AUC. Oracles come from the Therapeutics Data Commons via PyTDC (MIT); the ZINC database is free to use.

What is known

• Best in the PMO paper's table: SynNet at 0.969 ± 0.004, then Graph GA at 0.964 ± 0.012 and DoG-Gen at 0.948 ± 0.001. REINVENT, the best method overall across the 23 oracles, reaches 0.945 ± 0.007 here.
• Later work on the same protocol reports higher: Genetic GFN at 0.974 ± 0.006 and Mol GA at 0.950 ± 0.004 (five runs, 10,000 calls).
• A random screen of ZINC 250k scores 0.545, so the oracle is easy to satisfy but not trivially so.
• This oracle is close to saturated, and the headroom left is in sample efficiency rather than in the final score.

What would count as progress

• A mean AUC top-10 above 0.969 on drd2 from five runs under the official protocol, reported with the per-run values and the standard deviation. To beat the field rather than the 2022 table, the number to pass is 0.974 (Genetic GFN).
• The run logs, which record every scored molecule in order, so the AUC and the oracle-call count can be recomputed rather than trusted.
• Code in the repository's method layout, so the result can be re-run on the other 22 oracles: a method that wins on one oracle and loses elsewhere is the pattern PMO was built to expose.

References

1. W. Gao, T. Fu, J. Sun and C. W. Coley, "Sample Efficiency Matters: A Benchmark for Practical Molecular Optimization", NeurIPS 2022 Datasets and Benchmarks (2022). https://arxiv.org/abs/2206.12411
2. M. Olivecrona, T. Blaschke, O. Engkvist and H. Chen, "Molecular de-novo design through deep reinforcement learning", Journal of Cheminformatics 9, 48 (2017). https://doi.org/10.1186/s13321-017-0235-x
3. W. Gao, R. Mercado and C. W. Coley, "Amortized Tree Generation for Bottom-up Synthesis Planning and Synthesizable Molecular Design", ICLR 2022 (2021). https://arxiv.org/abs/2110.06389
4. H. Kim, M. Kim, S. Choi and J. Park, "Genetic-guided GFlowNets for Sample Efficient Molecular Optimization", NeurIPS 2024 (2024). https://arxiv.org/abs/2402.05961
5. A. Tripp and J. M. Hernández-Lobato, "Genetic algorithms are strong baselines for molecule generation" (2023). https://arxiv.org/abs/2310.09267

Source

Imported from PMO (mol_opt), the drd2 oracle, curated by Wenhao Gao, Tianfan Fu, Jimeng Sun and Connor W. Coley. Open according to the source as of 2026-10-20. The statement above is written in our own words; see the source for the original wording, history and discussion.