OdorSpace is the current status of a long-time effort in our lab to link odorant structure to odorant perception. This started with identifying a systematic link between the physical organization of the odor world and the ensuing perception of odorant pleasantness (Khan et al., 2007), a link also reflected in the neural representation of smell (Haddad et al., 2010).
We refined this computational framework to generate more precise predictions of odorant pleasantness from structure (Snitz et al., 2019), and initial predictions of pairwise odorant perceptual similarity from structure (Snitz et al., 2013). Recently, we have furthered this framework such that we can not only predict, but can also design odorants from structure.
Our initial resolution allowed us to reach at such creations as Olfactory White (Weiss et al., 2012), and our current and far-improved resolution allowed us to create olfactory metamers, namely odorant formulas containing non-overlapping sets of molecules that nevertheless smell exactly the same (Ravia et al., 2020). With this in hand, one can mix a novel version of any known target odor. For example, if you have the composition of the smell of "rose", you can mix a new "rose" that will smell the same, despite sharing no components with the original rose. In OdorSpace you can now also generate any novel formula, and we can tell you what it will smell like. The ability to generate compositions that smell like other smells mimics how the human olfactory system works (Yeshurun et al., 2009): a content-addressable memory system. Thus, we think that OdorSpace will provide the foundation for digitizing smell. We hope OdorSpace can help you in your olfaction research.