Research map / Learning theory and algorithms / Combinatorial and search optimization
APEX: Approximate-but-exhaustive search for ultra-large combinatorial synthesis libraries
Aryan Pedawi, Jordi Silvestre-Ryan, Bradley Worley, Darren J. Hsu, Kushal S Shah, Elias Stehle, Jingrong Zhang, Izhar Wallach
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 |
|---|---|---|---|
| run top-k on GPU instead of CPU | APEX search · BRICS 10B, unconstrained, k=1,000,000 | runtime (s): 147.6 -> 21.2 | Improved |
| run top-k on GPU instead of CPU | APEX search · REAL 70B, Lipinski, k=10,000 | runtime (s): 3163.2 -> 186.1 | Improved |
| run top-k on GPU instead of CPU | APEX search · BRICS 12M, unconstrained, k=100,000 | runtime (s): 1.2 -> 1.2 | No clear effect |
| run top-k on GPU instead of CPU | APEX search · BRICS 10B, Lipinski, k=10,000 | runtime (s): 437.7 -> 13.9 | Improved |
| add Lipinski Rule of 5 constraints | APEX search on GPU · BRICS 10B, k=10,000 | runtime (s): 10.9 -> 13.9 | Hurt |
| add Lipinski Rule of 5 constraints | APEX search on CPU · BRICS 10B, k=10,000 | runtime (s): 130.7 -> 437.7 | Hurt |
| increase retrieved set from k=100,000 to k=1,000,000 | APEX search on GPU · BRICS 10B, unconstrained | runtime (s): 11.6 -> 21.2 | Hurt |
| increase library from 10B to 70B compounds | APEX search on GPU · unconstrained, k=100,000 | runtime (s): 11.6 -> 169.3 | Hurt |
| use APEX instead of Thompson sampling | APEX · unconstrained top-j recall, five largest reactions | recall | Improved |
| apply pretrained surrogate and factorizer zero-shot to Enamine REAL | APEX · Enamine REAL docking score enrichment | docking score enrichment | Mixed |
Closest papers by meaning
- AANet: Virtual Screening under Structural Uncertainty via Alignment and Aggregation (NeurIPS 2025 · similarity 0.86)
- WelQrate: Defining the Gold Standard in Small Molecule Drug Discovery Benchmarking (NeurIPS 2024 · similarity 0.85)
- CBGBench: Fill in the Blank of Protein-Molecule Complex Binding Graph (ICLR 2025 · similarity 0.84)
- CombiMOTS: Combinatorial Multi-Objective Tree Search for Dual-Target Molecule Generation (ICML 2025 · similarity 0.84)
- DeltaDock: A Unified Framework for Accurate, Efficient, and Physically Reliable Molecular Docking (NeurIPS 2024 · similarity 0.84)
- TurboHopp: Accelerated Molecule Scaffold Hopping with Consistency Models (NeurIPS 2024 · similarity 0.84)
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.
