Presentation Details
Introduction
.

Abstract


Predicting how chemicals give rise to human odor perception remains one of the most challenging problems in sensory science. While machine learning has begun to close the gap between molecular structure and perceptual experience, meaningful progress depends on integrating high-quality perceptual data, robust computational models, and biological insight. This work collectively addresses how olfaction can transition from a largely descriptive discipline to a predictive and mechanistically informed science.�A key focus is the generation of scalable, standardized odor quality data that can support modern modeling approaches. Efficient methods for capturing perceptual meaning at scale are essential to enable reliable structure: percept mappings and to ensure that predictions generalize across stimuli, concentrations, and contexts. Equally important is moving beyond isolated odorants to account for the complexity of real-world smells, including concentration effects and multi-component mixtures that define everyday olfactory experiences.�The integration of predictive models with receptor-level biology further strengthens the link between chemistry and perception, enabling not only improved prediction accuracy but also rational strategies for modifying odor experiences. Together, these efforts illustrate a convergent framework in which sensory methodology, artificial intelligence, and neurobiological understanding work in concert to advance digital representations of smell, support innovation in fragrance and odor control, and deepen fundamental insight into how humans perceive complex chemical environments.

No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.