International Journal of Computer Applications (0975 - 8887) Volume 186 - No.31, July 2024 A Review on Protein Function Prediction Methods using Protein-Protein Interaction Networks Saima Khan Faculty, Computer Science and Engineering University of Development Alternative Dhaka, Bangladesh Md. Abidur Rahman Khan Computer Science and Engineering University of Asia Pacific Dhaka, Bangladesh ABSTRACT Proteins are essential components of all living organisms, perform- ing a myriad of biological functions within living bodies and sys- tems. Understanding protein functions is crucial for researchers, as it enables the development of various evolutionary medicines, treat- ments, and other beneficial products. However, many protein func- tions remain unknown. Computational methods have gained pop- ularity over traditional physical experiments for predicting protein functions. These methods include approaches based on sequence and structure knowledge, gene expression data, and protein-protein interaction data. Notably, protein function prediction methods uti- lizing protein-protein interaction networks have yielded more sat- isfactory results compared to those using other attributes. Proteins rarely function in isolation; they typically operate in conjunction with their interacting partners. Numerous researchers have pro- posed and implemented various techniques for accurately predict- ing the functions of unknown proteins. This paper presents a com- prehensive review of various methods proposed and utilized by re- searchers for predicting protein functions using protein-protein in- teraction networks. The descriptions include essential tables and figures, accompanied by appropriate citations and references. The aim of this paper is to assist other researchers in understanding these techniques and to encourage the development of enhanced approaches for predicting protein functions through protein-protein interaction networks. General Terms Protein, prediction, support, wet lab, dry lab, accuracy Keywords Protein function prediction, protein-protein interaction network, common neighbor, majority, neighbor protein 1. INTRODUCTION Protein function prediction represents a pivotal area of inquiry within the realm of bioinformatics. Despite significant strides, many protein functions remain elusive to researchers. A compre- hensive understanding of protein functions holds promise for ad- dressing a myriad of biological and medical challenges. Traditional wet lab methodologies have historically served as the cornerstone for elucidating protein functions. However, such approaches are characterized by inherent limitations, including time intensiveness, financial burden, and technical complexity. Consequently, compu- tational methods have emerged as an increasingly favored alter- native for predicting protein functions. These computational tech- niques offer expedited and streamlined predictions, entail reduced financial expenditure, and necessitate diminished human interven- tion compared to conventional wet lab methodologies. Wet lab- based experiments involve direct manipulation and testing of bi- ological or chemical materials in a laboratory environment. These experiments necessitate the use of physical samples and reagents, often employing techniques like mixing, culturing, and measur- ing under controlled conditions. Dry lab-based experiment refers a form of research or study carried out without the conventional wet lab methods, which usually entail working with chemicals, biolog- ical specimens, or other physically manipulable materials. Rather, dry lab experiments focus on computational techniques, simula- tions, data analysis, and theoretical modeling. Several computational methods exist for forecasting protein func- tions, including sequence-based, structure-based, protein-protein interaction network-based, gene expression data-based, and path- way analysis from gene expression data-based methods [1]. How- ever, the limited availability of protein structure data restricts the effective and widespread use of homology-based approaches in protein function prediction [1]. Many databases like SWISSPROT [25], DIP [26], NCBI [27], STRING [28], and PDB [29] are avail- able for using protein function prediction experiments. Among various computational methods for protein function predic- tion, leveraging protein-protein interaction networks emerges as a potent strategy for efficiently and swiftly predicting precise protein functions. Since proteins generally function through interactions with other proteins, protein-protein interaction networks provide crucial insights for predicting their functions. PPI networks are par- ticularly valuable for predicting protein functions because they pro- vide a comprehensive view of how proteins interact within a cell. By mapping these interactions, researchers can infer the roles of unknown proteins based on their interaction partners and network positions. Various computational techniques are applied to analyze PPI networks, including network clustering, machine learning, and data integration methods. 27