A graph neural network trained on molecular structure predicted novel, never-smelled molecules' odor descriptions well enough to beat the average individual human panelist, passing what the team called an 'odor Turing test.'

11:34panel. So let's compare that to our model. So the question is basically you know the odor touring test is if you want to make your panel better would you rather add another person or would you rather add the predictions of a model right um and it turned out our model predictions were better than any one individual panelist on average in the panel meaning we've passed a not during

11:57test like our model predictions were human quality which was pretty cool. Um that was the first thing and what we did with the neural network is we cracked it open and we looked at what's called the embedding layer which is a a part of the neural network that uh basically turns the inputs into a vector that's that is the map right and then that map is what

12:17we kind of can slice up in regions and use for classification. So this region of the map is vanilla this region of the map is redberry etc. Um without that you actually can't do that classification problem. So that embedding turned out if you do the engineering right it just kind of needs to be around 300 dimensions to work really well which is like suspicious but you know just

12:37suggestive right I can't make any claims um but uh that embedding had a ton of beautiful structure in it and that seemed to be a first candidate for a continuous predictive map of smell and we called it the principal odor map and that's been foundational to what we've built at the company since then. Before we get to that structure, you mentioned a graph neuron that was the fundamental

13:01architecture here and what did the the nodes and the edges in the graph represent? >> Yeah, great question. So like if you were doing machine learning on a social network graph, the nodes would be people and the edges would be relationships between people like friendships and and then it might be one very large graph of like Facebook or Twitter. uh in our case uh every uh graph is a molecule and the

13:27nodes of the molecule are the atoms. So it might be a carbon, it might be a nitrogen, it might be a sulfur or an oxygen. Um and the edges in the graph are the bonds and that might be a single bond, a double bond, a triple bond. And the graphs aren't very big because molecules that have a smell aren't very big. If they were huge, they actually