Recent advances in predictive models and the availability of perceptual data have enabled the prediction of odor perceptions from molecular structures for pure odorants. However, accurately modeling the perception of olfactory mixtures remains a significant challenge. The 2024 DREAM Olfaction Challenge focused on predicting the discriminability of olfactory mixtures. Building on this foundation, the 2025 DREAM Olfaction Challenge aims to advance the field by predicting the perceptual characteristics of olfactory mixtures. This challenge features a newly collected dataset of over 650 mixtures, each composed of 2, 3, 5, or 10 components from a pool of more than 145 monomolecular odorants. The perceptual data were gathered from a cohort of ≥15 trained panelists (two replicates) per stimulus using the rate-all-that-apply (RATA) method with a 51-term odor lexicon. Challenge participants will be tasked with developing models to predict the odor quality of these mixtures in the form of 51 semantic descriptors. The challenge will introduce new modeling opportunities for understanding the complex interactions between odorants and their perceptual outcomes, ultimately advancing our ability to predict and manipulate olfactory experiences.
