Code for project Geometric Deep Learning Models for Prediction of Protein Binding with Quadruplexes

Abstract

This paper focuses on one of the subtasks of bioinformatics - predicting the binding of proteins to G-quadruplex structures. It is well-known that proteins bind to nucleic acids and are essential for the regulation of cell growth and development. Initially, proteins were known to bind to duplex DNA, but there are proteins that interact with G-quadruplex structures of DNA and/or RNA and are involved in various biological functions, such as replication, transcription, genetic recombination, and other cellular activities. Investigating the interactions between proteins and nucleic acids that form G-quadruplex structures could be of great importance for various diseases, including cancer, psychiatric, and neurological diseases. Therefore, the ability to accurately predict their interaction could be beneficial for the discovery and development of new drugs and treatments. In this work, different ways of processing the nucleotide chain and protein sequence are considered and depth and machine learning models for predicting protein binding and G-quadruplex structure are implemented. The G4IPDB dataset was used as data.