Zhou, Yingyao, Cox, Jiayi, Zhou, Bin, Zhu, Steven, Zhong, Yang and Spraggon, Glen (2024) Afpdb – an efficient structure manipulation package for AI protein design. Bioinformatics.

Abstract

Motivation: The advent of AlphaFold and other protein Artificial Intelligence (AI) models has transformed protein design. To maximize the likelihood of success, the AI-driven protein design process can create thousands of design candidates. A typical AI design workflow involves handling large-scale structure file read/write operations, performing structure alignment, measuring deviations, standardizing chain/residue labels, extracting residues, identifying mutations, and automating visualization generations. Existing programming packages fall short of meeting these new requirements. To address this gap, we developed the Afpdb package.

Results: The Afpdb package, built upon AlphaFold’s NumPy architecture, significantly accelerates protein structure computations. Leveraging the intuitive contig syntax proposed by RFDiffusion, Afpdb streamlines residue and atom selection. Afpdb augments Biopython and other macromolecular structure frameworks with a suite of methods commonly used in protein AI design but are not readily available elsewhere. Additionally, it seamlessly integrates PyMOL’s visualization capabilities. In summary, Afpdb can enhance productivity in universal protein structure manipulation tasks within the structural biology community.

Item Type:	Article
Keywords:	Python, protein structure manipulation, AI protein design, AlphaFold.
Date Deposited:	19 Nov 2024 00:45
Last Modified:	19 Nov 2024 00:45
URI:	https://oak.novartis.com/id/eprint/54441