Computational gastronomy: ML and DL predictors for Umami peptides

Abstract

Umami is an important widely-used taste component of food seasoning. Umami peptides are spe- cific structural peptides endowing foods with a favorable umami taste. Laboratory approaches used to identify umami peptides are time-consuming and labor-intensive, which are not feasible for rapid screening. Taste is crucial in driving food choice and preference. Umami is one of the basic tastes defined by characteristic deliciousness and mouthfulness that it imparts to foods. Various models have been shown to predict umami taste using feature encodings derived from traditional molecular descriptors to transformer based feature embeddings. Over the course of the semester I implemented and explored the existing papers in the domain, applied and tested unique feature embeddings, evaluated machine learning and deep learning frameworks. I used the feature embeddings of Mol2Vec to extract a 300D feature vector, which was tested and gave a maximum accuracy of 92.1%.