Pose Reproduction, Cross Docking, Database Enrichment, and Reverse Docking Benchmarks for DOCK6, Vina, and AutoDock 4 with Emphasis on Approved Drugs

Abstract High-quality test sets to evaluate docking software help the community understand strengths and weaknesses of different methods and programs. In this work, we introduce SB2025, a calculation-ready set of 1,244 protein-ligand complexes derived from the PDB, curated using literature-based ligand protonation states, and freely available in DOCK6, AutoDock Vina, and AutoDock 4 (AD4) formats. SB2025 includes 322 unique proteins, 17 major protein classes, and ligands with a wide range of properties and flexibilities, making it a broad and challenging test set. FARMA2025, a specific subset of SB2025, contains 165 approved drugs bound to their pharmacological targets and closely mirrors the distribution of FDA drug target classes for small molecules through 2023 (r2 = 0.72). Comprehensive benchmarking was subsequently performed to evaluate pose reproduction (PR), cross-docking (CD), database enrichment (DE), and reverse docking (RD) under standardized conditions. Averaged PR success rate across triplicate runs for DOCK6 was slightly improved compared to Vina (followed by AD4) using both the large SB2025 test set and the more drug-focused FARMA2025 test set. Conversely, CD matrix success rates for AD4 were more favorable than DOCK6 or Vina across 11 protein families. The programs all yielded substantial early DE; however, the number of active ligands in common between the programs within the top 1% of the database, with one exception, was surprisingly low. Challenging RD tests using FARMA2025, which gauge the ability to correctly rank the cognate structure best, indicate that generating the correct ligand pose is necessary but not sufficient. SB2025, FARMA2025, and example scripts and protocols for benchmarking are available at ringo.ams.stonybrook.edu (downloads) and github.com/rizzolab.