Biological Attributes in Classifying Protein Complex Membership

Proteins are crucial for various biological processes, often functioning as part of protein complexes that enable complex cellular activities. While traditional network-based approaches predict entire protein complexes by identifying dense subgraphs in protein-protein interaction (PPI) networks, this study focuses on predicting individual protein membership in complexes using their biological attributes. A binary classification model was developed, leveraging features derived from protein sequence properties, bag-of-words features, Kidera property factors, and amino acid physicochemical and biochemical attributes. Comprehensive feature selection methods were employed to enhance model performance. The proposed approach utilizes a Support Vector Classifier (SVC) and Random Forest (RF) model for classification, leveraging protein-specific characteristics as features. Experimental results demonstrate the SVC model’s potential for accurately classifying protein membership, providing valuable insights into protein functions and their roles within complexes.