Binding Mode Matters: Hotspot-Aware Drug Discovery via Explorative Preferences
Organizations: Institute of Automation, Chinese Academy of Sciences · School of Artificial Intelligence, University of Chinese Academy of Sciences · Zhongguancun Academy & Zhongguancun Institute of Artificial Intelligence · Shanghai Institute for Advanced Immunochemical Studies, ShanghaiTech University ·
Abstract
Discovering hit molecules requires not just high binding affinity, but also the identification of diverse binding modes, which are critical for experimental assays. Existing generative approaches have predominantly relied on optimizing a scalar docking score, obscuring the distinct contributions of key binding determinants. To this end, we introduce a paradigm shift by formulating target-based drug design as a multi-objective exploration task, where each objective explicitly corresponds to enhancing interactions with a specific hotspot. Here, we introduce BindMol, a novel generative framework driven by a customized multi-objective reinforcement learning algorithm. By incorporating explorative preferences during training, our approach efficiently uncovers molecules with diverse and desirable binding profiles. Empirical results demonstrate that BindMol facilitates the discovery of high-affinity compounds characterized by both structural novelty and diverse binding modes. Validated across target-based drug discovery and multi-property optimization tasks, our approach provides a versatile paradigm for goal-oriented drug discovery.
Figures & tables
| Method | Target protein | ||||
|---|---|---|---|---|---|
| parp1 | fa7 | 5ht1b | braf | jak2 | |
| REINVENT [ 40 ] | 0 0.480 ( 0.344) | 0 0.213 ( 0.081) | 0 2.453 ( 0.561) | 0 0.127 ( 0.088) | 0 0.613 ( 0.167) |
| Graph GA [ 24 ] | 0 4.811 ( 1.661) | 0 0.422 ( 0.193) | 0 7.011 ( 2.732) | 0 3.767 ( 1.498) | 0 5.311 ( 1.667) |
| MORLD [ 25 ] | 0 0.047 ( 0.050) | 0 0.007 ( 0.013) | 0 0.880 ( 0.735) | 0 0.047 ( 0.040) | 0 0.227 ( 0.118) |
| HierVAE [ 26 ] | 0 0.553 ( 0.214) | 0 0.007 ( 0.013) | 0 0.507 ( 0.278) | 0 0.207 ( 0.220) | 0 0.227 ( 0.127) |
| RationaleRL [ 27 ] | 0 4.267 ( 0.450) | 0 0.900 ( 0.098) | 0 2.967 ( 0.307) | 0 0.000 ( 0.000) | 0 2.967 ( 0.196) |
| Method | Target protein | ||||
|---|---|---|---|---|---|
| parp1 | fa7 | 5ht1b | braf | jak2 | |
| REINVENT [ 40 ] | 0 -8.702 ( 0.523) | -7.205 ( 0.264) | 0 -8.770 ( 0.316) | 0 -8.392 ( 0.400) | 0 -8.165 ( 0.277) |
| Graph GA [ 24 ] | -10.949 ( 0.532) | -7.365 ( 0.326) | -10.422 ( 0.670) | -10.789 ( 0.341) | -10.167 ( 0.576) |
| MORLD [ 25 ] | 0 -7.532 ( 0.260) | -6.263 ( 0.165) | 0 -7.869 ( 0.650) | 0 -8.040 ( 0.337) | 0 -7.816 ( 0.133) |
| HierVAE [ 26 ] | 0 -9.487 ( 0.278) | -6.812 ( 0.274) | 0 -8.081 ( 0.252) | 0 -8.978 ( 0.525) | 0 -8.285 ( 0.370) |
| RationaleRL [ 27 ] | -10.663 ( 0.086) | -8.129 ( 0.048) | 0 -9.005 ( 0.155) | No hit found | 0 -9.398 ( 0.076) |
| Method | Target protein | ||||
|---|---|---|---|---|---|
| parp1 | fa7 | 5ht1b | braf | jak2 | |
| REINVENT [ 40 ] | 0 44.2 ( 15.5) | 23.2 ( 6.6) | 138.8 ( 19.4) | 18.0 ( 2.1) | 0 59.6 ( 0 8.1) |
| MORLD [ 25 ] | 0 0 1.4 ( 0 1.5) | 0 0.2 ( 0.4) | 0 22.2 ( 16.1) | 0 1.4 ( 1.2) | 0 0 6.6 ( 0 3.7) |
| HierVAE [ 26 ] | 0 0 4.8 ( 0 1.6) | 0 0.8 ( 0.7) | 0 0 5.8 ( 0 1.0) | 0 3.6 ( 1.4) | 0 0 4.8 ( 0 0.7) |
| RationaleRL [ 27 ] | 0 61.3 ( 0 1.2) | 0 2.0 ( 0.0) | 312.7 ( 0 6.3) | 0 1.0 ( 0.0) | 199.3 ( 0 7.1) |
| FREED [ 59 ] | 0 34.8 ( 0 4.9) | 21.2 ( 4.0) | 0 88.2 ( 13.4) | 34.4 ( 8.2) | 0 59.6 ( 0 8.2) |
| Method | Benchmark | Average | ||||||
|---|---|---|---|---|---|---|---|---|
| Amlodipine | Fexofenadine | Osimertinib | Perindopril | Ranolazine | Sitagliptin | Zaleplon | ||
| REINVENT [ 40 ] | 0.608 | 0.752 | 0.806 | 0.511 | 0.719 | 0.006 | 0.325 | 0.532 |
| MolDQN [ 62 ] | 0.230 | 0.431 | 0.636 | 0.125 | 0.018 | 0.000 | 0.002 | 0.206 |
| Graph GA [ 24 ] | 0.622 | 0.731 | 0.799 | 0.503 | 0.670 | 0.330 | 0.305 | 0.566 |
| Graph MCTS [ 24 ] | 0.385 | 0.522 | 0.655 | 0.219 | 0.694 | 0.117 | 0.165 | 0.394 |
| DoG-Gen [ 4 ] | 0.489 | 0.640 | 0.706 | 0.422 | 0.601 | 0.015 | 0.073 | 0.421 |
| Metric | Method | PARP1 | FA7 | 5HT1B | BRAF | JAK2 |
|---|---|---|---|---|---|---|
| IFP clusters | GEAM | 138.67 | 61.67 | 8.33 | 20.33 | 47.33 |
| BindMol | 192.00 | 94.33 | 11.67 | 35.00 | 77.33 | |
| Entropy | GEAM | 2.93 | 1.85 | 0.24 | 0.68 | 1.52 |
| BindMol | 3.45 | 2.37 | 0.32 | 1.11 | 2.31 | |
| Effective modes | GEAM | 18.73 | 6.36 | 1.27 | 1.97 | 4.57 |
| BindMol | 31.40 | 10.71 | 1.37 | 3.03 | 10.10 |
Appendix figures & tables11 assets
Supplementary material from the paper’s appendix.
Appendix
| Target | Method | IFP clusters | IFP entropy | Effective modes |
|---|---|---|---|---|
| PARP1 | REINVENT | |||
| FREED | ||||
| RationaleRL | ||||
| GEAM | ||||
| BindMol | ||||
| FA7 | REINVENT |
| Method | PARP1 | FA7 | 5HT1B | BRAF | JAK2 |
|---|---|---|---|---|---|
| GPI | |||||
| C-MORL | |||||
| PreCo | |||||
| BindMol |
| Method | PARP1 | FA7 | 5HT1B | BRAF | JAK2 |
|---|---|---|---|---|---|
| Full union (including intersection) | |||||
| Non-overlapping residues |
| Method | IFP clusters | IFP entropy | Effective modes |
|---|---|---|---|
| BindMol -h | |||
| BindMol |
| Metric | Target | 10% | 25% | 50% | 75% | 100% |
|---|---|---|---|---|---|---|
| Novel hit ratio (%) | BRAF | |||||
| JAK2 | ||||||
| Novel top-5% docking score | BRAF | |||||
| JAK2 |
| Method | Benchmark | ||||||
|---|---|---|---|---|---|---|---|
| Amlodipine | Fexofenadine | Osimertinib | Perindopril | Ranolazine | Sitagliptin | Zaleplon | |
| REINVENT [ 40 ] | 303.7 | 343.3 | 452.3 | 318.3 | 253.3 | 398.3 | 275.3 |
| Graph GA [ 24 ] | 258.7 | 333.3 | 270.3 | 278.7 | 364.7 | 306.3 | 272.7 |
| STONED [ 39 ] | 303.7 | 330.3 | 301.3 | 301.0 | 316.7 | 326.3 | 280.3 |
| GEAM-static [ 35 ] | 412.0 | 397.7 | 315.3 | 318.0 | 256.7 | 233.0 | 267.0 |
| GEAM [ 35 ] | 424.0 | 502.0 | 435.0 | 377.3 | 295.3 | 257.0 | 336.0 |
Molecular Generation
Molecular generation develops computational methods to create chemical structures and conformations for discovery, optimization, and structure elucidation.
See all 253 papers