Graph Neural Networks for Guiding Chemical Synthesis of Metal–Organic Frameworks

This project will explore how graph neural networks (GNNs) can be used to probabilistically predict successful crystallization conditions in the reticular synthesis of novel metal–organic frameworks (MOFs). MOFs are rich chemical structures with important applications in carbon capture and storage, water harvesting from air, hydrogen storage, and catalysis—areas central to addressing global energy and sustainability challenges.

The goal of this project is to develop deep learning models, specifically graph neural networks that are potentially pre-trained on large molecular databases, to overcome limitations of earlier models (random forests and Gaussian processes with hand-crafted features) that fail to generalize to unseen molecular components. The GNN models will be integrated into a closed-loop AI–robot system at UC Berkeley performing high-throughput synthesis experiments. In the context of this system a central research question is how to provide accurate probabilistic predictions, which are crucial for balancing exploration and exploitation in the search for new MOFs.

Prerequisites

• Machine learning, ideally some experience in deep learning and probabilistic models
• Python and the scipy stack
• Some background or interest in learning about the application domain of chemistry