A principal odor map unifies diverse tasks in olfactory perception

2023/09/01 by Brian Lee, Emily J. Mayhew, Benjamín Sánchez-Lengeling +11 · 1 voice · 19 citations
Agricultural and Biological Sciences · Engineering · Neuroscience · #Advanced Chemical Sensor Technologies #Insect Pheromone Research and Control #Olfactory and Sensory Function Studies

paper · doi:10.1126/science.ade4401

openalex publication_date 2023/09/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/30

Abstract

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.

Citations

• Predicting human olfactory perception from chemical features of odor molecules

Cited by

• Hierarchies of smell: structuring the molecular odor space using semantic taxonomies and machine learning
• Beyond Molecular Structure: Critically Assessing Machine Learning for Designing Organic Photovoltaic Materials and Devices
• AI for Sustainable Future Foods
• Artificial Intelligence Empowered New Materials: Discovery, Synthesis, Prediction to Validation
• Storing Sensor Events in the Interconnection Strength of Conducting Polymer Dendrites
• Benchmark for Assessing Olfactory Perception of Large Language Models
• Odors as cognitive constructs: history of odor classification and attempts to map odor percepts to physical and chemical parameters
• Response to Doty 2025, predicting odor from structure is useful
• Response to Comments of Pellegrino et al.
• Ranking over Regression for Bayesian Optimization and Molecule Selection
• Mouthfeel assessment of alternative proteins: what we know so far?
• Rotational Sampling: A Plug-and-Play Encoder for Rotation-Invariant 3D Molecular GNNs
• CheMixHub: Datasets and Benchmarks for Chemical Mixture Property Prediction
• Diffusion Graph Neural Networks and Dataset for Robust Olfactory Navigation in Hazard Robotics
• Position: Olfaction Standardization is Essential for the Advancement of Embodied Artificial Intelligence
• What makes a scent trigger a memory? A cognitive decomposition of odor-evoked retrieval
• SmellNet: A Large-scale Dataset for Real-world Smell Recognition
• Design Principles From Natural Olfaction for Electronic Noses
• Multi-Hierarchical Fine-Grained Feature Mapping Driven by Feature Contribution for Molecular Odor Prediction

Discussions

• In 2023, machine learning models predicting odor character from molecular structure outperformed the median trained human panelist on a prospective test of 400 molecules. These models are now genuine [bsky, 2 points, 1 comments]

Related

• Machine Learning for Scent: Learning Generalizable Perceptual Representations of Small Molecules
• From Molecules to Mixtures: Learning Representations of Olfactory Mixture Similarity using Inductive Biases
• Olfactory Label Prediction on Aroma-Chemical Pairs
• Molecular Odor Prediction Based on Multi-Feature Graph Attention Networks
• Predicting human olfactory perception from chemical features of odor molecules