Mofasa: A Step Change in Metal-Organic Framework Generation

2025-12-01

AI Summary

Current MOF generation models exhibit insufficient performance, hindering rational design for critical applications such as atmospheric water harvesting, carbon capture, gas storage, and catalysis. To address this, we propose the first all-atom latent diffusion model that jointly samples atomic types, 3D coordinates, and lattice vectors in an end-to-end manner—eliminating reliance on manual assembly or predefined topologies. The model simultaneously generates metal nodes, organic linkers, and global topology, supporting complex MOFs with up to 500 atoms. It overcomes key bottlenecks of conventional generative paradigms, achieving substantial improvements in structural diversity, validity, and crystallographic quality over state-of-the-art methods. Concurrently, we release MofasaDB, an open-source database containing hundreds of thousands of annotated MOF structures, and an interactive visualization and search platform. These resources collectively advance AI-driven materials design and enable reproducible, community-wide research.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Mofasa is an all-atom latent diffusion model with state-of-the-art performance for generating Metal-Organic Frameworks (MOFs). These are highly porous crystalline materials used to harvest water from desert air, capture carbon dioxide, store toxic gases and catalyse chemical reactions. In recognition of their value, the development of MOFs recently received a Nobel Prize in Chemistry. In many ways, MOFs are well-suited for exploiting generative models in chemistry: they are rationally-designable materials with a large combinatorial design space and strong structure-property couplings. And yet, to date, a high performance generative model has been lacking. To fill this gap, we introduce Mofasa, a general-purpose latent diffusion model that jointly samples positions, atom-types and lattice vectors for systems as large as 500 atoms. Mofasa avoids handcrafted assembly algorithms common in the literature, unlocking the simultaneous discovery of metal nodes, linkers and topologies. To help the scientific community build on our work, we release MofasaDB, an annotated library of hundreds of thousands of sampled MOF structures, along with a user-friendly web interface for search and discovery: https://mofux.ai/ .

Problem

• Generates novel Metal-Organic Framework structures using a diffusion model
• Samples atomic positions, types, and lattice vectors for large systems
• Enables simultaneous discovery of metal nodes, linkers, and topologies

Innovation

• All-atom latent diffusion model for MOF generation
• Simultaneously samples positions, atom-types, lattice vectors
• Avoids handcrafted assembly algorithms for discovery