Generation of drug-like molecules with spatial conformations driven by expected gene expression signature and target receptor

Abstract

A student of the KN-414 group, Prymachenko Maksym, completed the bachelor's qualification work. Topic "Generation of drug-like molecules with spatial conformations driven by expected gene expression signature and target receptor." The work aims to obtain a bachelor's degree in the specialty 122 "Computer Science." The object of research is the de novo design of small molecules using neural networks. The research aims to develop a neural network for generating drug-like molecules with spatial conformations driven by expected gene expression signatures and target receptors. Shape-based features are used to represent a molecule. Together with compressed data on gene expression, they form a shared latent space. From this latent space, having one part, another is generated. The molecules themselves are generated from the generated forms to the text representation. As a result of the bachelor’s qualification works, was developed and trained a neural network for the de novo design of drug-like molecules. The total volume of work: 59 pages, 25 figures, 27 references.