FRaPPÉ: Fragrance Raw Materials Performance Prediction Explained

Abstract

The evaporation and deposition behavior of perfume raw materials in consumer products is governed by their physicochemical interactions across complex matrices, yet current predictive approaches remain empirical and poorly suited to esterquat-based fabric softeners. This work presents a QSPR framework for predicting perfume raw materials’ activity coefficients in the bottle phase and activities on wet and dry fibers, using experimentally measured headspace data for 98 perfume raw materials. A feature selection strategy is developed to identify interpretable features. Three modeling workflows─independent, parallel, and serial─are constructed to explore different information transfer approaches between touchpoints. The serial workflow achieves the most robust performance, particularly for fiber-based predictions under limited sample sizes, demonstrating that sequential propagation of physicochemical information enhances generalization. SHAP analyses reveal hydrophobicity and topological complexity as dominant descriptors governing perfume raw materials partitioning and evaporation, yielding an interpretable tool for fragrance ingredient screening and product development.