Metal binding site alignment enables network-driven discovery of recurrent geometries across sequence-divergent proteins and drug off-targets
Vetle Simensen, Eivind Almaas
Abstract
Metal-binding sites (MBSs) are critical determinants of protein stability and biological function, yet methods for comparing their local binding environments lag behind those for whole-structure alignment. Here, we represent MBSs as atomic point clouds surrounding bound metal ligands and align them with a fine-tuned iterative closest point algorithm. Applying this framework to a redundancy-reduced collection of MBSs derived from all metalloproteins in the Protein Data Bank (PDB), we perform pairwise alignments across 23,342 sites to construct a similarity network of metal-binding environments.
Introduction
Metal-binding proteins are ubiquitous in biological systems, where bound metals contribute to a multitude of functions essential for cellular processes [1,2]. Broadly, the bound metal play either of two roles: structural or functional. Structurally, the metal ligands are critical for the proper folding and stabilization of proteins into their native three-dimensional structure [1]. Functionally, they frequently facilitate biochemical catalysis in metalloenzyme active sites, commonly serving as redox centers or conduits for electron transfer, or as structural scaffolds that orient substrates and stabilize transition states [2].
Methods
Dataset construction and extraction of metal-binding site point clouds
We obtain protein structures from the RCSB PDB via its REST API (accessed 2024-07-20) [23]. We query the PDB for entries containing chemical components with the target metal atoms listed in Table 1. Candidate MBSs are then defined from individual metal atom coordinates parsed from the associated PDBx/mmCIF files. Thus, metal-containing compounds are not treated as complete MBS centers; rather, each retained metal atom instance provides one candidate MBS and is subject to the following quality filter: We retain only crystallographic structures with reported overall resolution <3.0 Å.
Results
MBS dataset overview and impact of redundancy filtering
We assemble an initial metalloprotein dataset by parsing PDB entries containing any of the queried metals (Table 1) and extracting, for each metal ligand, an MBS point cloud from surrounding protein atoms (see Methods). Because closely related structures are frequently deposited multiple times (often with multiple biological assemblies), we apply a redundancy reduction procedure that clusters highly similar sites within protein- and metal-matched groups and retains representative point clouds.
Discussion
In this work, we extend and formalize the ICP-based framework of Raymond et al. [21] for aligning point cloud representations of MBSs. Whereas their study focused on a small set of dimanganese proteins and defined sites as all atoms within a fixed-radius sphere of the metal ion, we curate a comprehensive collection of all MBSs in the PDB and adopt a fixed-cardinality representation.
Acknowledgments
The authors would like to express their gratitude to Emil Karlsen for his assistance with the setup and execution of the all-to-all pairwise point cloud alignment on our in-house compute cluster. We also thank Gaston Courtade for constructive and valuable feedback on the manuscript.
Citation: Simensen V, Almaas E (2026) Metal binding site alignment enables network-driven discovery of recurrent geometries across sequence-divergent proteins and drug off-targets. PLoS Comput Biol 22(8): e1014636. https://doi.org/10.1371/journal.pcbi.1014636
Editor: Amar Singh, KU: The University of Kansas, UNITED STATES OF AMERICA
Received: March 3, 2026; Accepted: July 27, 2026; Published: August 28, 2026
Copyright: © 2026 Simensen, Almaas. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data and code underlying this study are available at the MBSNetwork GitHub repository via https://doi.org/10.5281/zenodo.20793348. This includes the code used for the analyses and figures, together with the data needed to reproduce the results.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
