A principal odor map unifies diverse tasks in olfactory perception

• 成果类型: Article
• 署名作者: Lee, Brian K.; Mayhew, Emily J.; Sanchez-Lengeling, Benjamin; Wei, Jennifer N.; Qian, Wesley W.; Little, Kelsie A.; Andres, Matthew; Nguyen, Britney B.; Moloy, Theresa; Yasonik, Jacob; Parker, Jane K.; Gerkin, Richard C.; Mainland, Joel D.; Wiltschko, Alexander B.
• 署名单位: Alphabet Inc.; Google Incorporated; Monell Chemical Senses Center; Michigan State University; University of Illinois System; University of Illinois Urbana-Champaign; University of Reading; Arizona State University; Arizona State University-Tempe; University of Pennsylvania
• 刊物名称: SCIENCE
• ISSN/ISSBN: 0036-10043
• DOI: 10.1126/science.ade4401
• 发表日期: 2023-09-01
• 页码: 999-1006
• 关键词: neural-networks

摘要

Mapping molecular structure to odor perception is a key challenge in olfaction. We used graph neural networks to generate a principal odor map (POM) that preserves perceptual relationships and enables odor quality prediction for previously uncharacterized odorants. The model was as reliable as a human in describing odor quality: On a prospective validation set of 400 out-of-sample odorants, the model-generated odor profile more closely matched the trained panel mean than did the median panelist. By applying simple, interpretable, theoretically rooted transformations, the POM outperformed chemoinformatic models on several other odor prediction tasks, indicating that the POM successfully encoded a generalized map of structure-odor relationships. This approach broadly enables odor prediction and paves the way toward digitizing odors.