A team at the Daegu Gyeongbuk Institute of Science and Technology (DGIST) has published a detailed roadmap for an artificial olfactory system that pairs metal-organic frameworks (MOFs) with AI to distinguish tens of thousands of odors. The review, published July 13, 2026 in Progress in Materials Science (impact factor 42.9, top 0.7% in JCR), lays out a path from sensor material design to AI-driven odor recognition, with direct applications in disease diagnosis, environmental monitoring, and smart agriculture.
Electronic noses, or e-noses, detect odor molecules through multiple sensors and then use AI to learn the resulting signal patterns. Current sensor materials often struggle with selectivity, slow response times, and rigid operating conditions. MOFs-porous crystals formed from metal ions and organic linkers-can be tuned to adsorb specific molecules, and they operate at room temperature under low power. The roadmap systematically organizes the Research into three MOF-based material groups: pristine MOFs, MOF-composites, and MOF-derivatives, each offering different trade-offs in sensitivity, stability, and selectivity.
How the human nose inspires artificial olfaction
The design mirrors the biological nose. Humans identify a vast range of smells using only about 400 receptor types because each odor triggers a unique combination of receptors-a principle called combinatorial coding. The DGIST team applied the same logic: an array of MOF sensors, each with a distinct response profile, generates a pattern of signals that a machine learning model then classifies. The group showed how deep learning can extract subtle features from these sensor arrays, moving beyond simple yes/no detection toward nuanced interpretation of complex odor mixtures.
Bridging materials science and AI pattern recognition
Prof. Hyuk-Jun Kwon, who led the research, said, "MOFs provide a virtually unlimited materials library that can be designed to exhibit different responses to different odors, much like human olfactory receptors. This paper is significant in that it bridges the gap between materials research and AI-based odor recognition research while presenting a roadmap for the development of intelligent electronic noses tailored to specific applications." The study's alignment with AI for Science & Research reflects a broader push to use machine learning to accelerate materials discovery and sensor design.
The roadmap categorizes the enabling technologies not just by material type but also by the AI techniques used. Simple machine learning classifiers can separate known odor classes, while deep neural networks can learn to pick out features that humans might not predefine. The combination of tunable MOF arrays and data-driven analysis, the authors argue, is what will push e-nose systems beyond laboratory demonstrations into real-world deployment.
Why this matters for Science and Research professionals
For researchers in materials science, analytical chemistry, or AI, the review offers a structured survey of a field that sits squarely at the intersection of hardware and software. It identifies which MOF architectures currently pair best with which AI models, and it flags underexplored areas-such as long-term sensor drift correction and few-shot learning for new odors-that present immediate opportunities for grant proposals and new collaborations. The roadmap also makes the case that the next generation of e-nose systems will be built by teams that combine deep domain knowledge in both porous materials and neural network design, a signal to hiring committees and funding agencies that interdisciplinary work is no longer optional.
Your membership also unlocks:
