Navigating the Fragrance space Via Graph Generative Models And Predicting Odors

TL;DR

This paper introduces a novel framework combining graph generative models and machine learning to generate, validate, and predict odors of molecules, significantly advancing fragrance discovery and chemical space exploration.

Contribution

It presents an integrated approach for molecule generation, odor prediction, and interpretability, with high accuracy and practical tools for fragrance research.

Findings

ROC AUC score of 0.97 for odor prediction

Effective correlation between physicochemical features and odor likeliness

Open-source code and models for community use

Abstract

We explore a suite of generative modelling techniques to efficiently navigate and explore the complex landscapes of odor and the broader chemical space. Unlike traditional approaches, we not only generate molecules but also predict the odor likeliness with ROC AUC score of 0.97 and assign probable odor labels. We correlate odor likeliness with physicochemical features of molecules using machine learning techniques and leverage SHAP (SHapley Additive exPlanations) to demonstrate the interpretability of the function. The whole process involves four key stages: molecule generation, stringent sanitization checks for molecular validity, fragrance likeliness screening and odor prediction of the generated molecules. By making our code and trained models publicly accessible, we aim to facilitate broader adoption of our research across applications in fragrance discovery and olfactory research.