Research map / Probabilistic and scientific modeling / Molecules and proteins
Guiding Deep Molecular Optimization with Genetic Exploration
Sungsoo Ahn, Junsu Kim, Hankook Lee, Jinwoo Shin
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use GEGL instead of the strongest reported baseline | PenalizedLogP optimization · PenalizedLogP | objective: 27.22 -> 31.40 | Improved |
| Use GEGL instead of the strongest reported baseline | Similarity-constrained molecular optimization · PenalizedLogP, δ=0.4 | objective: 3.98 -> 7.87 | Improved |
| Use GEGL instead of the strongest reported baseline | Similarity-constrained molecular optimization · PenalizedLogP, δ=0.6 | objective: 2.49 -> 4.43 | Improved |
| Use GEGL rather than the compared methods | GuacaMol benchmark · 20 GuacaMol tasks | number of tasks with highest score: highest score for 19 out of 20 tasks | Improved |
| Apply post-hoc filtering to generated molecules | GEGL · GuacaMol benchmark with filtering | task score: 0.837 -> 0.552 | Mixed |
| Remove the genetic expert policy (ablation) | GEGL apprentice policy · Sitagliptin MPO and Zaleplon MPO | GuacaMol score | Hurt |
| Remove the neural apprentice policy and retain expert-policy updates (ablation) | GEGL · Sitagliptin MPO and Zaleplon MPO | GuacaMol score | Hurt |
| Use GB-GA with GEGL's genetic-expert hyperparameters (ablation) | GEGL genetic-exploration setup · Sitagliptin MPO and Zaleplon MPO | GuacaMol score | Hurt |
| Train the apprentice on genetic-expert samples as well as its own high-reward samples (ablation) | GEGL · Sitagliptin MPO and Zaleplon MPO | GuacaMol score | Improved |
Closest papers by meaning
- Genetic-guided GFlowNets for Sample Efficient Molecular Optimization (NeurIPS 2024 · similarity 0.87)
- Searching for High-Value Molecules Using Reinforcement Learning and Transformers (ICLR 2024 · similarity 0.83)
- Neural Genetic Search in Discrete Spaces (ICML 2025 · similarity 0.83)
- Molecular Optimization Model with Patentability Constraint (AAAI 2024 · similarity 0.83)
- Molecule Design by Latent Prompt Transformer (NeurIPS 2024 · similarity 0.83)
- Enhancing Ligand Validity and Affinity in Structure-Based Drug Design with Multi-Reward Optimization (ICML 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
