PREDISCENT - Numerical models for predicting smell and emotion from chemical features

Project Leader: Jérôme Golebiowski, ICN

Project Partners: Institut de Chimie de Nice et Expressions Parfumées

France is one of the major protagonists in the field of flavor and fragrance. Innovation through the identification of new smelling compounds however mostly relies on serendipity. Indeed, for scientists, establishing a relationship between the structure of a molecule and its chemosensory properties is a long-standing challenge. Intuitively, perfumers' know-how implicitly suggests that such relationships exist and the challenge lies in their decoding. Besides perception, these experts build their compositions to communicate emotions which are even harder to quantify. The reasons for such a difficulty relies on i/ the extremely complex biological system involved in chemosensory perceptions, and ii/ potential genetic and/or cultural variations amongst the human population.

Can we learn a computer how to smell or even feel relaxed? Can we initiate a new field of "computational behavioral neuroscience"? We need to deeply explore how numerical models and high-performance computing can open the area of the prediction of smell or emotions.

Our project aims to connect machine learning algorithms and molecular modeling with properties measured through sensory analysis and physiology experiments on panels of human individuals.

Project start and end dates: January 2018 - July 2019