FP-Zernike: An Open-source Structural Database Construction Toolkit for Fast Structure Retrieval

Abstract

The release of AlphaFold2 has sparked a rapid expansion in protein model databases. Efficient protein structure retrieval is crucial for the analysis of structure models, while measuring the similarity between structures is the key challenge in structural retrieval. Although existing structure alignment algorithms can address this challenge, they are often time-consuming. Currently, the state-of-the-art approach involves converting protein structures into three-dimensional (3D) Zernike descriptors and assessing similarity using Euclidean distance. However, the methods for computing 3D Zernike descriptors mainly rely on structural surfaces and are predominantly web-based, thus limiting their application in studying custom datasets. To overcome this limitation, we developed FP-Zernike, a user-friendly toolkit for computing different types of Zernike descriptors based on feature points. Users simply need to enter a single line of command to calculate the Zernike descriptors of all structures in customized datasets. FP-Zernike outperforms the leading method in terms of retrieval accuracy and binary classification accuracy across diverse benchmark datasets. In addition, we showed the application of FP-Zernike in the construction of the descriptor database and the protocol used for the Protein Data Bank (PDB) dataset to facilitate the local deployment of this tool for interested readers. Our demonstration contained 590,685 structures, and at this scale, our system required only 4–9 s to complete a retrieval. The experiments confirmed that it achieved the state-of-the-art accuracy level. FP-Zernike is an open-source toolkit, with the source code and related data accessible at https://ngdc.cncb.ac.cn/biocode/tools/BT007365/releases/0.1, as well as through a webserver at http://www.structbioinfo.cn/.

1. Introduction

Proteins, as the building blocks of all living systems, fold into specific three-dimensional (3D) configurations and perform corresponding biological functions. To understand the mechanism of protein action at the molecular level, it is necessary to accurately predict the 3D structure of proteins. AlphaFold2 [1] has made significant strides in protein structure prediction, achieving comparable prediction accuracy with experimental methods through a well-designed deep neural network and greatly reducing the prediction time. This breakthrough suggests that the protein model structure database will grow at an amazing speed. The latest AlphaFold database release contains over 200 million entries, highlighting the urgent need for an efficient and accurate method to measure protein structural similarity.

Structural alignment is the most direct method to measure the similarity between structures. There are two main types of structural alignment methods: coordinate-based methods and surface-based methods. Coordinate-based methods, dating back to the 1970s [2], focus on the superimposition of structures based on atomic coordinate information. In the following decades, various schemes have been proposed and improved, such as combinatorial extension (CE) [3], DALI [4], RNA-align [5], TM-align [6], and US-align [7]. Since the nature of the protein surface is crucial to the study of protein–protein (RNA, ligand) interactions, structure alignment schemes based on the protein surface have been proposed, such as gmfit [8], ZEAL [9], and iterative closest point (ICP) [10]. However, these alignment methods are often time-consuming. For example, a standard alignment software (gmfit) takes ~0.71 s to complete a structural alignment and ~5 days for a structure retrieval (~590,000 alignments).