Abstract
SignificanceWhat makes a molecule have a smell? This simple question represents a significant gap in our understanding of olfaction. To answer it, we trained models to predict whether molecules were odorous based on molecular characteristics and noted which characteristics were needed to make correct predictions. We found that molecules with sufficient volatility and hydrophobicity are generally odorous, which suggests that reaching olfactory receptors is the dominant barrier for prospective olfactory stimuli. Based on these criteria, there are billions of molecules that are likely odorous but have never been smelled. We can now recognize an odorous molecule from its structure, and this guides us to discover new classes of odorants and include all types of odorants in our study of smell.
Original language | English (US) |
---|---|
Pages (from-to) | e2116576119 |
Journal | Proceedings of the National Academy of Sciences of the United States of America |
Volume | 119 |
Issue number | 15 |
DOIs | |
State | Published - Apr 12 2022 |
Keywords
- machine learning
- odor space
- olfaction
- physical transport
ASJC Scopus subject areas
- General