PDB Manipulation Suite (PDBMS): A browser-based toolkit for PDB file manipulation, molecule building and biophysical analysis, easily extendable via reasoning models

We present here the PDB Manipulation Suite (PDBMS, https://pdbms.altervista.org/ ), a modular hub of web apps and platforms designed for the seamless creation, edition, and analysis of biomolecular coordinate files, all online and client-side i.e. without any user files even passing by any server. The suite is organized into specialized sub-platforms: (i) PDBMS-tools for rapid structure analysis and editing, supporting residue renumbering, sequence extraction, B-factor hacking, CIF/PDB conversion, basic calculations such as distances or solvent accessibility, and various means for mapping data; (ii) Web-PackMol for the assembly of complex molecular systems; and (iii) Web-MDA and the Conformational Ensemble Inspector for the deep analysis and inspection of single structures as well as ensembles coming from molecular dynamics trajectories, BioEmu predictions of conformational landscapes, docking poses, NMR models, etc., all assisted by a novel form of AI copilot that we introduce here. The various tools at the PDBMS facilitate inference of structural details, dynamics, and interactions across molecules and across time or ensemble indexes; others aid in the visualization of various kinds of data including AlphaFold 3 confidence scores and outputs from modern NMR spectroscopy pipelines. Since the source code for all tools at the PDBMS is inherently open and was developed largely with AI reasoning models at Google’s AI Studio, it is extremely easy to modify or add new functionalities at will. We demonstrate this by providing customized tools to assist the editing and visualization of outputs from NMR-specific software like NMRtist and Sparky; and we also demonstrate how the AI agent built into the Web-MDA app can provide new functionalities on request including modification of the app’s own GUI. By providing a “zero-footprint” environment available out of the box with no installs, no downloads, no registration, and trivially clonable code for custom local modifications and reuse, we hope the PDBMS will facilitate work in structural bioinformaticians, structural biology and molecular modeling online but preserving absolute data confidentiality, extreme portability across devices and operating systems, and maximal flexibility. We also hope these tools and ideas for in-software AI use will inspire others to develop further smart, client-only scientific software.