The CASP 16 Experimental Protein-Ligand Datasets

Area of Science:

• Structural Biology
• Drug Discovery
• Computational Chemistry

Background:

• CASP (Critical Assessment of protein Structure Prediction) experiments traditionally focus on protein structure prediction.
• The integration of pharmaceutical discovery projects into CASP16 marks a significant expansion into real-world drug development challenges.

Purpose of the Study:

• To present experimental protein-ligand datasets for CASP16, specifically for drug discovery targets.
• To establish benchmarks for evaluating computational methods in protein-ligand pose and affinity prediction for pharmaceutical applications.

Main Methods:

• Assembly and characterization of protein-ligand complexes for four drug target proteins: human chymase, human cathepsin G, human autotaxin, and SARS-CoV-2 main protease.
• Collection of over 200 co-crystal structures (<2.7 Å resolution) and binding affinity measurements for ~160 compounds (nanomolar to high micromolar range).
• Detailed description of experimental approaches including protein production, crystallization, and binding assay development.

Main Results:

• Creation of benchmark datasets featuring pharmaceutically relevant protein-ligand complexes.
• Inclusion of systems with challenging characteristics, such as specific protein surfaces, ligand properties, and active site conformations.
• Data generation from actual drug discovery projects, providing a realistic testbed.

Conclusions:

• The presented datasets are crucial for advancing computational drug discovery by providing realistic benchmarks.
• These datasets will facilitate the assessment of computational methods' performance on targets relevant to pharmaceutical research.
• The study highlights the successful integration of drug discovery projects into the CASP framework.