MidZPPI: a sequence-based approach for predicting multi-label protein-protein Iinteractions in midnight-zone of protein sequence alignments

Abstract

The essential building blocks of life are known as protein, which performs a wide range of tasks in our bodies while the most important role of the proteins is interacting with each other and performing cellular activities. However, the challenging part is to predict the Protein-Protein Interactions (PPI) as they hold a complicated nature and in order to predict these interactions accurately, advanced computational approaches are required. This thesis inspects sequence-based approaches for the prediction of MultiLabel Protein-Protein Interaction (PPI) types amidst the Midnight Zone of protein sequence alignments. These proteins in Midnight Zone are hard to explore, has similar structures rarely with lack of enough features which makes it computationally challenging. By using advanced Deep Learning (DL) methods and Deep Neural Networks (DNN) on protein sequences, our study determines to perform the prediction and classification of these Protein-Protein Interactions (PPI) in the Midnight Zone of protein sequence alignments. This paper proposes a novel MidZPPI (Midnight Zone Protein-Protein Interaction) prediction model, aims to label Protein-Protein Interaction (PPI) types properly on our prepared MHS62k dataset which consists of proteins ofMidnight Zone sequence alignments. This model uses a subgraph approach based on the Protein Protein Interaction (PPI) types and uses Graph Neural Networks (GNN), Graph Convolutional Networks (GCN) and Autoencoders (AE) on this. This paper also provides a wide range of knowledge about the functionalities of other zones in protein sequence alignments. Our thesis promotes the field of bioinformatics, which is expanding every day by a variety of methods and techniques that not only aims to predict Protein Protein Interaction (PPI) types, but also contributing in discovering new ways to produce new drugs and medicines, and also the exploration of critical diseases.