Abstract:Motivation: Protein-protein interactions are important for many biological processes. Theoretical understanding of the structurally determining factors of interaction sites will help to understand the underlying mechanism of protein-protein interactions. Taking advantage of advanced mathematical methods to correctly predict interaction sites will be useful. Although some previous studies have been devoted to the interaction interface of protein monomer and the interface residues between chains of protein dimers, very few studies about the interface residues prediction of protein multimers, including trimers, tetramer and even more monomers in a large protein complex. As we all know, a large number of proteins function with the form of multibody protein complexes. And the complexity of the protein multimers structure causes the difficulty of interface residues prediction on them. So, we hope to build a method for the prediction of protein tetramer interface residue pairs. Results: Here, we developed a new deep network based on LSTM network combining with graph to predict protein tetramers interaction interface residue pairs. On account of the protein structure data is not the same as the image or video data which is well-arranged matrices, namely the Euclidean Structure mentioned in many researches. Because the Non-Euclidean Structure data can't keep the translation invariance, and we hope to extract some spatial features from this kind of data applying on deep learning, an algorithm combining with graph was developed to predict the interface residue pairs of protein interactions based on a topological graph building a relationship between vertexes and edges in graph theory combining multilayer Long Short-Term Memory network. First, selecting the training and test samples from the Protein Data Bank, and then extracting the physicochemical property features and the geometric features of surface residue associated with interfacial properties. Subsequently, we transform the protein multimers data to topological graphs and predict protein interaction interface residue pairs using the model. In addition, different types of evaluation indicators verified its validity.

Triplet encoded sequence based membrane protein classification using BiLSTM

DeepTrio: a Ternary Prediction System for Protein-Protein Interaction Using Mask Multiple Parallel Convolutional Neural Networks

Prediction of membrane protein types in a hybrid space.

ProtienCNN‐BLSTM: An efficient deep neural network with amino acid embedding‐based model of protein sequence classification and biological analysis

Identifying Membrane Protein Types Based on Lifelong Learning With Dynamically Scalable Networks

Protein-Protein Interactions Prediction Based on Bi-directional Gated Recurrent Unit and Multimodal Representation

Protein sequence classification using natural language processing techniques

Macromolecule Classification Based on the Amino-acid Sequence

An Efficient Deep Learning Approach for DNA-Binding Proteins Classification from Primary Sequences

An hierarchical artificial neural network system for the classification of transmembrane proteins

Hybrid framework for membrane protein type prediction based on the PSSM

DeepPLM_mCNN: An approach for enhancing Ion Channel and Ion Transporter Recognition by Multi-Window CNN based on features from Pre-trained Language Models

Feature Assisted bi-directional LSTM Model for Protein-Protein Interaction Identification from Biomedical Texts

Machine learning in computational modelling of membrane protein sequences and structures: from methodologies to applications

Prediction of Multi-Type Membrane Proteins in Human by an Integrated Approach.

Tetramer protein complex interface residue pairs prediction with LSTM combined with graph representations

Deducing high-accuracy protein contact-maps from a triplet of coevolutionary matrices through deep residual convolutional networks

Protein Classification using Machine Learning and Statistical Techniques: A Comparative Analysis

Enhanced Protein Structural Class Prediction Using Effective Feature Modeling and Ensemble of Classifiers

MultiPredGO: Deep Multi-Modal Protein Function Prediction by Amalgamating Protein Structure, Sequence, and Interaction Information

Research on DNA-Binding Protein Identification Method Based on LSTM-CNN Feature Fusion