Predicting Bitterness Inducing Peptides Through Traditional Supervised Machine Learning

Bitter peptides can affect both the taste and the healthiness of a given food product. Identification of bitter peptides can allow researchers to develop healthy and palatable foods and medicines. The program Orange by Demsar et al. (2013) was used to analyze a dataset of peptides and develop machine learning models to predict if a given peptide causes bitterness. The gradient boosting tree and random forest models performed the most optimally out of all models. This can be used as a basis for improving on bitter peptide detection through machine learning in the future.