 ESP, enzyme substrate prediction, is a machine learning model developed to predict enzyme substrate pairs from a set of known enzymes and their corresponding substrates. The model was trained using data augmentation techniques and random sampling of non-substrates, resulting in an accuracy of over 91 percent. This model can be used to identify potential substrates for any given enzyme, regardless of its family or type. This article was authored by Alexander Kroll, Sir Hossra Ranjan, Martin K. M. Enquist, and others.