[Paper Review] PyMOLfold: Interactive Protein and Ligand Structure Prediction in PyMOL

Colby T. Ford, Arif Ullah+2 more|ArXiv.org|Feb 1, 2025

Machine Learning in Bioinformatics3 citations

TL;DR

PyMOLfold is an open-source PyMOL plugin that integrates AI-based protein structure prediction and ligand placement directly in PyMOL, using models like ESM3, Boltz-1, and Chai-1 to predict protein tertiary structures from sequences and optionally place ligands from SMILES strings within the predicted structures.

ABSTRACT

PyMOLfold is a flexible and open-source plugin designed to seamlessly integrate AI-based protein structure prediction and visualization within the widely used PyMOL molecular graphics system. By leveraging state-of-the-art protein folding models such as ESM3, Boltz-1, and Chai-1, PyMOLfold allows researchers to directly predict protein tertiary structures from amino acid sequences without requiring external tools or complex workflows. Furthermore, with certain models, users can provide a SMILES string of a ligand and have the small molecule placed in the protein structure. This unique capability bridges the gap between computational folding and structural visualization, enabling users to input a primary sequence, perform a folding prediction, and immediately explore the resulting 3D structure within the same intuitive platform.

Motivation & Objective

• Enable protein tertiary structure prediction directly from amino acid sequences inside PyMOL.
• Integrate state-of-the-art folding models (e.g., ESM3, Boltz-1, Chai-1) into a single visualization workflow.
• Allow optional placement of small-molecule ligands via SMILES strings within the predicted protein structures.
• Provide an interactive, user-friendly workflow that bridges computational folding and structural visualization in PyMOL.

Proposed method

• Develop a flexible, open-source PyMOL plugin that interfaces with AI-based folding models.
• Utilize protein folding models such as ESM3, Boltz-1, and Chai-1 to predict tertiary structures from sequences.
• Support input of SMILES strings to place ligands within the predicted protein structure.
• Deliver an interactive visualization experience inside PyMOL for immediate exploration of results.

Experimental results

Research questions

• RQ1 Can AI-based folding models be directly integrated into PyMOL to predict protein structures from sequences without external tools?
• RQ2 Is it possible to place a ligand in the predicted protein structure using a SMILES string within the same PyMOL environment?
• RQ3 Does the integrated workflow streamline the transition from sequence to folded structure and structural visualization?
• RQ4 What are the practical limitations and usability considerations of combining folding models with PyMOL visualization?
• RQ5 How do different models (ESM3, Boltz-1, Chai-1) compare within the PyMOLfold plugin in terms of performance and results?

Key findings

• The PyMOLfold plugin enables direct protein tertiary structure prediction from amino acid sequences within PyMOL.
• Users can optionally provide a SMILES string to place a ligand in the protein structure, bridging folding and visualization.
• The plugin leverages state-of-the-art models such as ESM3, Boltz-1, and Chai-1 to perform folding predictions.
• The approach emphasizes seamless integration and interactive exploration of results inside PyMOL.