AROMMA: Unifying Olfactory Embeddings for Single Molecules and Mixtures

2026-01-27

AI Summary

Existing public olfactory datasets are limited in scale and treat single molecules and mixtures separately, hindering the learning of generalizable odor representations. This work proposes the first unified embedding space that leverages a chemical foundation model to encode individual molecules and introduces a permutation-invariant attention-based aggregator to model binary mixtures, effectively capturing their asymmetric interactions. To address missing labels, the approach employs knowledge distillation and class-aware pseudo-labeling, enabling end-to-end alignment of odor representations for both single molecules and mixtures. Evaluated across multiple olfactory benchmarks, the method achieves state-of-the-art performance, with AUROC improvements up to 19.1%, and demonstrates significantly enhanced cross-dataset generalization capability.

Technology Category

• Machine Learning: Representation Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Public olfaction datasets are small and fragmented across single molecules and mixtures, limiting learning of generalizable odor representations. Recent works either learn single-molecule embeddings or address mixtures via similarity or pairwise label prediction, leaving representations separate and unaligned. In this work, we propose AROMMA, a framework that learns a unified embedding space for single molecules and two-molecule mixtures. Each molecule is encoded by a chemical foundation model and the mixtures are composed by an attention-based aggregator, ensuring both permutation invariance and asymmetric molecular interactions. We further align odor descriptor sets using knowledge distillation and class-aware pseudo-labeling to enrich missing mixture annotations. AROMMA achieves state-of-the-art performance in both single-molecule and molecule-pair datasets, with up to 19.1% AUROC improvement, demonstrating a robust generalization in two domains.

Problem

• olfactory embeddings
• single molecules
• mixtures
• unified representation
• odor perception

Innovation

• unified olfactory embedding
• attention-based aggregator
• chemical foundation model
• knowledge distillation
• asymmetric molecular interactions