AFIR identifies various reaction pathways that also include irrelevant molecular structures. However, when paired with ChemOntology, it can identify all relevant reaction pathways faster while ignoring irrelevant molecular structures. Credit: ACS Catalysis (2025). DOI: 10.1021/acscatal.5c06298

Chemical reactions are the breaking and forming of chemical bonds, which are fundamental to the creation of new technologies. Inevitably, discovering and developing new chemical reactions is a time-intensive process of trial and error.

To support chemical reaction discovery, a research team from WPI-ICReDD, led by Professor Masaharu Yoshioka and Assistant Professor Pinku Nath, have developed ChemOntology—a new artificial intelligence system that rapidly explores and analyzes chemical reactions with human-like intuition.

The research is published in ACS Catalysis.

How ChemOntology enhances reaction discovery

ChemOntology is a chemical knowledge classification system that formalizes human chemical reasoning into a machine-interpretable framework that dynamically integrates with the Artificial Force Induced Reaction (AFIR) computational method previously developed by Professor Satoshi Maeda, director of WPI-ICReDD.

"AFIR identifies reaction pathways by systematically generating new molecular structures and evaluating their energies. However, because this process considers far more possibilities than chemists would typically consider realistic, it requires substantial computational time and costs," explained Dr. Yuriko Ono.

When paired with ChemOntology, AFIR can intuitively recognize which chemical bonds are likely to participate in reactions and is able to distinguish chemically reasonable structures from unrealistic ones. Effectively, this technology has successfully programmed a (human) chemist’s intuition and does not rely on training datasets, an advantage over machine-learning.

For proof of concept, researchers analyzed the classical Heck reaction, which contains complex reaction mechanisms and several major products, using AFIR and ChemOntology. Credit: ACS Catalysis (2025). DOI: 10.1021/acscatal.5c06298

Proof of concept and research impact

As a proof of concept, the researchers analyzed a classical and well-known process, a Heck reaction, using AFIR alone and in combination with ChemOntology. The Heck reaction mechanism involves several complex intermediates that can produce up to three major reaction products over 10 possible reaction steps, making it an ideal reaction to evaluate computational performance.

"AFIR alone identified only partial pathways for the Heck reaction. Amazingly, AFIR with ChemOntology successfully discovered all the pathways in half the computational time," said Assistant Professor Nath.

ChemOntology represents a new class of chemical discovery tool and an explicitly knowledge-driven and training-free artificial intelligence framework. It directly embeds human chemical reasoning into the reaction search process. This new tool not only reduces computational costs but makes the analyses more meaningful.

This research was a collaborative effort made possible with the insights of computational chemists, Dr. Yuriko Ono, Professor Yu Harabuchi, Professor Satoshi Maeda, and Professor Tetsuya Taketsugu, along with those of experimental chemist Professor Yasunori Yamamoto.

More information: Pinku Nath et al, ChemOntology: A Reusable Explicit Chemical Ontology-Based Method to Expedite Reaction Path Searches, ACS Catalysis (2025). DOI: 10.1021/acscatal.5c06298

Citation: AI mimics human-like intuition to explore and analyze chemical reactions (2025, December 22) retrieved 22 December 2025 from https://phys.org/news/2025-12-ai-mimics-human-intuition-explore.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.