A comprehensive dataset of 32 million pentapeptide structures for high-throughput virtual screening

Area of Science:

• Computational chemistry and structural biology.
• Bioinformatics and data science.

Background:

• Small peptides are crucial in biology but their conformational flexibility hinders structure-based studies.
• High-throughput screening and structure-based design of peptides are challenging due to conformational complexity.

Purpose of the Study:

• To generate and release an exhaustive open dataset of three-dimensional (3D) structures for all canonical amino-acid pentapeptides.
• To provide a computational resource that facilitates virtual screening, method benchmarking, and machine learning in peptide research.

Main Methods:

• An automated computational workflow was developed using UCSF ChimeraX, Reduce, and RDKit.
• The workflow generated up to 10 conformers for each of the 3,200,000 unique pentapeptide sequences.
• Structures were validated using RMSD analysis, Ramachandran quality assessment, and comparison with Protein Data Bank data.

Main Results:

• A dataset comprising 32,000,000 peptide conformers was generated and made publicly available.
• The dataset is distributed with an index for efficient programmatic access and subset selection.
• The computational workflow, built on open-source software, is adaptable for generating conformer libraries for other short peptides.

Conclusions:

• The comprehensive pentapeptide conformer dataset removes the need for researchers to generate such data from scratch.
• This resource significantly advances peptide design, protein engineering, and computational drug discovery.
• The open nature of the dataset and workflow promotes wider accessibility and application in the scientific community.