Predicting the taste of molecules is of critical importance in the food and beverages, flavor, and pharmaceutical industries for the design and screening of new tastants. In this work, we have built deep learning models to classify sweet, bitter, and umami molecules— the three basic tastes whose sensation is mediated by G protein-coupled receptors. An extensive dataset containing 1938 bitter, 2079 sweet, and 98 umami tastants was curated from existing literature. We analyzed the chemical characteristics of the molecules, with special focus on the presence of different functional groups. A deep neural network model based on molecular descriptors and a graph neural network model were trained for taste prediction. The class imbalance due to fewer umami molecules was tackled using special sampling techniques, such that the classwise metrics for all the three taste classes are optimized. Both models show comparable performance during evaluation, but the graph-based model can learn task-specific representations from the molecular structure without requiring handcrafted features. We further explain the deep neural network predictions using Shapley additive explanations and connect them to the physics of tastant-receptor binding. This study develops an in-silico approach to classify molecules based on their taste by leveraging the recent progress in deep learning, which can serve as a powerful tool for tastant design.

Calculating the free energy of drug permeation across membranes carries great importance in pharmaceutical and related applications. Traditional methods, including experiments and molecular simulations, are expensive and time-consuming, and existing statistical methods suffer from low accuracy. In this work, we propose a hybrid approach that combines molecular dynamics simulations and deep learning techniques to predict the free energy of permeation of small drug-like molecules across lipid membranes with high accuracy and at a fraction of the computational cost of advanced sampling methods like umbrella sampling. We have performed several molecular dynamics simulations of molecules in water and lipid bilayers to obtain multidimensional time-series data of features. Deep learning architectures based on Long Short-Term Memory networks, attention mechanisms, and dense layers are built to estimate free energy from the time series data. The prediction errors for the test set and an external validation set are much lower than that of existing data-driven approaches, with R2 of the best model around 0.99 and 0.82 for the two cases. Our approach reduces the time required for free energy calculations by an order of magnitude. This work presents an attractive option for high-throughput virtual screening of molecules based on their membrane permeabilities, demonstrates the applicability of language processing techniques in biochemical problems, and suggests a novel way of integrating physics with statistical learning to great success

Predicting the taste of molecules is of critical importance in the food and beverages, flavor, and pharmaceutical industries for the design and screening of new tastants. In this work, we have built deep learning models to classify sweet, bitter, and umami molecules— the three basic tastes whose sensation is mediated by G protein-coupled receptors. An extensive dataset containing 1466 bitter, 1764 sweet, and 238 umami tastants was curated from existing literature. We analyzed the chemical characteristics of the molecules, with special focus on the presence of different functional groups. A deep neural network model based on molecular descriptors and a graph neural network model were trained for taste prediction. The class imbalance due to fewer umami molecules was tackled using special sampling techniques. Both models show comparable performance during evaluation, but the graph-based model can learn task-specific representations from the molecular structure without requiring handcrafted features. We further explain the deep neural network predictions using Shapley additive explanations. Finally, we demonstrated the applicability of the models by screening bitter, sweet, and umami molecules from a large food database. This study develops an in-silico approach to classify molecules based on their taste by leveraging the recent progress in deep learning, which can serve as a powerful tool for tastant design.