Models that learn the language of molecules and how to use them to predict their properties

Abstract

Deep learning techniques are becoming increasingly important in all fields of science, but they have enormous potential, especially in biomedicine, to facilitate the diagnosis of diseases or accelerate drug discovery. In this workshop there will be an exhibition of what neural networks are, how they work and what applications they have. We will explain what language models are and how they can be used to predict properties of molecules (proteins, nucleic acids and drugs) and we will walk through different techniques to evaluate the generated models. The course will feature a case study in which attendees will be able to retrain a language model to predict small molecule toxicity and compare different evaluation methods.