Color has RGB. Sound has MP3s. Smell has nothing: no encoding, no way to work from first principles. For centuries, the only way to understand a scent was to smell it. We’ve spent the last year building the tools to decode it. Today, Patina is announcing our first Scent Foundation Model, understanding scent at the biological level, in order to develop never-before-smelled molecules and replace the vulnerable supply of natural ingredients.
All senses are built on receptors: proteins that convert a stimulus into a signal the brain can interpret. Our eyes have just 3 types of color receptors. This simplicity gave us RGB; every screen, camera, and digital image is built on it.
Our noses have close to 400 types of olfactory receptors, each responding to different molecular features. There is no compression scheme for smell, no “frequency” of rose that can be transmitted electronically. Smell requires actual molecules binding to actual receptor proteins.
Previous researchers used descriptors like “floral,” “woody,” “green” to describe scent, but these are cultural and inconsistent across languages. Perfumers anchor on known reference materials, but these cover a small subset of olfactory space.
The next step is receptors. The receptor code works universally across humans, operating beneath culture at the level of shared biology. Each molecule activates a distinct combination of these ~400 receptors, and it is this pattern that gives rise to what we perceive. No one has mapped it across the full set: until now.
Today, we are announcing Sense1: our first Scent Foundation Model and the first foundation model for molecular sensory perception. Our model enables a compression algorithm for scent: analogous to the RGB system for displaying colors. Sense1 is best-in-class, strongly outperforming traditional computational approaches at modeling molecule-receptor interaction.
Below is Sense1’s prediction for a single molecule: vanillin. Each cell is one of ~400 olfactory receptors: color coded by family, with brightness representing the receptor’s activation. In the same way a QR code uniquely represents binary computer data, this Scent Receptor (SR) Code uniquely represents a specific scent impression.
This key breakthrough of our model in the field of scent digitization allows for scent compression. Because the same SR Code can be created with different mixtures of molecules, any scent can now be encoded in a universal, standard set of primary scents.
Modeling the real world, however, means moving beyond single molecules. Natural essences and artistic fragrances are complex blends: rose oil alone contains over 300 chemical components. Below is Sense1’s prediction of what happens when you smell a rose: the activation pattern across all ~400 receptors.
Completely new scent molecules. We are hunting for the primary colors of scent: the palette from which any smell can be recreated. These new molecules will be safer, less skin-sensitizing, and cover never-before-smelled sensory profiles. Imagine the scent-equivalent of neon, or non-spectral colors.
Faithful replication of natural materials. Plants record their environment and replay it to us as complex aromas: rose oil alone contains over 300 chemical components, and no synthetic reconstruction has matched it. Understanding the receptor code is the key to faithful replication, especially as climate change disrupts supply chains.
On-demand scent and flavor. The receptor code also opens the door to designing scents and flavors to specification: composing new sensory experiences the way a digital artist composes with color, or a musician uses a synthesizer.
Health and biology. Olfactory receptors are not only found in the nose: they have been discovered in skin, gut, brain, and nearly every other tissue in the human body. The next generation of fragrances will do more than just smell good: they will actively benefit skin health, neuroplasticity, or interact with other biological pathways in targeted ways.
We built Sense1 because it was a tool that needed to exist. We built it for the naturalists trying to replicate extracts that climate change is making scarce, for the flavorists that are working from intuition where they should have data, and for brands that spend years developing a perfume that will be duplicated in weeks. Patina enables faster development, access to molecules that didn’t exist before, and the ability to match ingredients that are harder to source every year.
Now we are building our portfolio of captive molecules and working with select perfumers, brands, flavorists, and fragrance houses who want to build on this new foundation.
Patina is building the sensory foundation model: a computational understanding of how molecules become smells, tastes, and biological signals. Founded by Sean Raspet, a flavor scientist and artist who has spent the last 12 years studying the molecular structure of scent, and Laura Sisson, a former Google engineer who has been building machine learning models for olfaction since before it was a field. Sense1 is our first release, focused on olfaction.
Learn more at patina.earth.