Search

Deliberate training of GNN for PPI

Author(s)

Xinyao Li*

Affiliation(s)

Xi’an Jiaotong-Liverpool University, Suzhou, China *Corresponding Author

Abstract

Proteins perform most functions in cells, regulate and control the activities of cells. As their functions, predicting their interactions tend to be importantly. However, traditional PPI prediction techniques still have some limitations. For example, molecular dynamics simulation requires a lot of time and manpower; Some models (homologous modeling) lack accuracy when dealing with large amounts of data. In order to solve these defects, this paper tends to combine GNN (Graph Neural Network) and GNN characteristics to predict and analyze protein interactions, and discuss some applications. And analyze its advantages and disadvantages. First, GNN can improve the forecasting ability of PPI. Through the data analysis of the existing PPI map, the structure of the protein was transformed into a clearer nodal form, and the potential relationship between proteins was analyzed. In addition, GNN can help analyze PPI from different sources for information integration and data collation, so as to predict the complex relationship network among proteins. These GNN-guided results can be applied to the study of disease mechanisms, as well as the discovery and optimization of drug targets, with significant implications for medical pharmacy. For example, in the study of neurodegenerative diseases (Alzheimer's disease), revealing the network characteristics of abnormal protein interactions provides a new perspective for the study of pathological mechanisms of the disease.

Keywords

Graph Neural Network (GNN); Protein-Protein Interaction

References

[1] Keskin O, Gursoy A, Ma B, et al. Principles of protein− protein interactions: what are the preferred ways for proteins to interact?[J]. Chemical reviews, 2008, 108(4): 1225-1244. [2] Ni D, Lu S, Zhang J. Emerging roles of allosteric modulators in the regulation of protein‐protein interactions (PPIs): A new paradigm for PPI drug discovery[J]. Medicinal research reviews, 2019, 39(6): 2314-2342. [3] Li H, Sun X, Cui W, et al. Computational drug development for membrane protein targets[J]. Nature Biotechnology, 2024, 42(2): 229-242. [4] Browne F, Zheng H, Wang H, et al. From Experimental Approaches to Computational Techniques: A Review on the Prediction of Protein‐Protein Interactions[J]. Advances in Artificial Intelligence, 2010, 2010(1): 924529. [5] Wu J. Introduction to convolutional neural networks[J]. National Key Lab for Novel Software Technology. Nanjing University. China, 2017, 5(23): 495. [6] Wu Z, Pan S, Chen F, et al. A comprehensive survey on graph neural networks[J]. IEEE transactions on neural networks and learning systems, 2020, 32(1): 4-24. [7] Yang J, Li Y, Wang G, et al. An End-to-End Knowledge Graph Fused Graph Neural Network for Accurate Protein-Protein Interactions Prediction[J]. IEEE/ACM Transactions on Computational Biology and Bioinformatics, 2024. [8] Xiang S, Zhu M, Cheng D, et al. Semi-supervised credit card fraud detection via attribute-driven graph representation[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2023, 37(12): 14557-14565. [9] Zhao S, Cui Z, Zhang G, et al. MGPPI: multiscale graph neural networks for explainable protein–protein interaction prediction[J]. Frontiers in Genetics, 2024, 15: 1440448. [10] Sherstinsky A. Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network[J]. Physica D: Nonlinear Phenomena, 2020, 404: 132306. [11] Jin M, Xue H, Wang Z, et al. ProLLM: protein chain-of-thoughts enhanced LLM for protein-protein interaction prediction[J]. bioRxiv, 2024: 2024.04. 18.590025.