How AI Is Decoding MOFs: From 36 Years of Nobel-Worthy Research to Generative Design

MOFs Win the 2025 Nobel Prize and Enter the AI Era

On October 8, 2025, the Nobel Committee awarded the Chemistry Prize to Prof. Susumu Kitagawa (Kyoto University), Prof. Richard Robson (University of Melbourne), and Prof. Omar Yaghi (UC Berkeley) for their seminal contributions to metal‑organic frameworks (MOFs). The award highlights MOFs’ huge cavities that can capture water from desert air, extract pollutants, or store hydrogen, underscoring their unprecedented functional potential.

36 Years of Chemical Puzzles: From Coordination Polymers to MOFs

In 1989, Richard Robson proposed the first three‑dimensional coordination polymer, linking metal nodes and organic ligands into periodic networks. His JACS paper “Infinite Polymeric Frameworks Consisting of Three‑Dimensionally Linked Rod‑like Segments” (co‑authored with Ben F. Hoskins) provided the first experimental evidence of the MOF concept.

Over the next fifteen years, Omar Yaghi and Kitagawa’s teams published a series of breakthrough papers in Nature and Science , establishing MOFs as a distinct porous‑material family and moving the field into a systematic expansion stage.

Kitagawa later introduced the concepts of “Flexible Frameworks” and “Breathing MOFs,” demonstrating reversible gas uptake and release, and turning MOFs from rigid pores into dynamic, responsive materials.

In 1999, Yaghi created the highly stable MOF‑5, showing that rational design could endow MOFs with new functions such as hydrogen storage. His “reticular chemistry” paradigm made crystal synthesis predictable. Concurrently, Mohamed Eddaoudi published “Metal‑Organic Frameworks from Design Strategies to Applications,” advancing high‑surface‑area MOFs.

Collaborations with Uppsala University produced the Zr‑based UiO series, which have become commercially viable, high‑temperature‑stable MOFs.

AI Meets MOFs: A Symbiotic Revolution

A bibliometric study by Yuzheng You’s group (South University of China) titled “Artificial Intelligence in Metal–Organic Frameworks from 2013 to 2024” showed a sharp rise in AI‑MOF publications after 2016, indicating growing research interest.

MOFs’ modular composition—metal nodes, organic linkers, and topological nets—creates a discrete, enumerable chemical space. As described in a JACS paper by KAIST, each component represents a separate variable (metal cluster/coordination, organic ligand scaffold, and topology), enabling systematic enumeration.

These variables allow graph neural networks to learn adsorption energies, thermal stability, and other properties directly from structure, while scalar descriptors (cell parameters, pore size, surface area) serve as labels for supervised learning.

Generative AI for MOF Design

Researchers from Tianjin University, the Chinese Academy of Sciences, and A*STAR reported that AI advances—large databases, deep learning architectures, generative models, and hybrid simulations—are transforming high‑performance MOF design, enabling accurate property prediction, automated structure generation, and large‑scale synthesis planning.

In 2024, KAIST and POSTECH introduced MOFFlow , a flow‑matching generative model that treats metal nodes and linkers as rigid bodies in SE space, reducing structural complexity and producing candidate frameworks.

In April 2025, Yaghi’s team at UC Berkeley launched MOFGen , an agentic AI system that combines a large‑language‑model “MOFMaster,” a linker generator, a diffusion‑based crystal generator (CrystalGen), quantum‑mechanics pre‑screening (QForge), synthesizability assessment (SynthABLE), distance validation (QHarden), and experimental verification (SynthGen) across seven coordinated steps.

Subsequent work from Peking University, Harvard, and Cambridge introduced a building‑block‑aware diffusion model that encodes individual molecular fragments and crystal topology, achieving high‑fidelity generation of thousands‑atom MOFs.

Experimental results show that the BBA MOF Diffusion model can generate novel MOFs with over a thousand atoms, breaking the “from‑scratch” limitation and offering a practical route to synthesizable high‑performance materials.

Outlook

The trajectory—from Robson’s coordination polymer prototype, through Yaghi’s reticular chemistry, to today’s AI‑driven generative pipelines—mirrors chemistry’s shift from empirical discovery to computational, data‑driven design. As generative AI, quantum computing, and high‑throughput experimentation converge, MOF research is poised to move from experience‑based discovery to fully automated, data‑centric material creation.