Odors Smell Like Their Components: A Linear Framework for Predicting Olfactory Mixture Perception

Abstract

Most smells, whether in food, perfume, or the environment, are complex mixtures of many odor molecules. A principled framework for modeling natural mixtures has been considered intractable because mixtures are widely assumed to generate nonlinear perceptual interactions that mirror nonlinear responses at the receptor and neural levels. We quantified perceptual interactions across 432 mixtures composed of 144 component odorants using a trained human panel, and nearly all mixtures were explained by a linear average of their component quality profiles. Within this linear modeling framework, averaging component profiles predicted mixture perception significantly better than the previous state-of-the-art and approached the ceiling set by measurement noise. Even mixtures previously reported to exhibit emergence were equally well predicted. These results suggest that the main challenge in modeling odor perception is characterizing individual components, rather than exhaustively mapping odor-odor interactions. Just as additive color mixing enabled colorimetry, linear mixing in olfaction opens the door to quantitative odorimetry, allowing mixtures to be predicted, reconstructed, and optimized computationally.