Comparative Analysis of Protein Complex Identification Methods Using Topological and Biological Information

Identifying the protein complexes improves knowledge in disease pathology and contributes to drug development and tailored illness treatments. The motivation behind the experiment is to find significant interactions between proteins, examine how mutant units can impair the activity of other proteins in PPINs, and discover ways to enhance the performance of protein complex detection methods by combining biological context information with topological data. This study compares some widely used protein complex detection techniques, including MCL, COACH, MCODE, IPAC, and DPCLUS, where there is a scope for incorporating biological information in addition to focusing on topological aspects. Predicted complexes are validated by comparing them with benchmark complexes that are known. The results are further analysed using the DAVID tool suite to find the impact of integrating gene expression data in predicting significantly enriched protein groups. In the case of human PPIs, the effectiveness percentages 86.1, 92.9, 84.6, 88.2, and 83.0, respectively, are shown by the above methods with an average of 10.52% increase when gene expression information was integrated with edge betweenness centrality scores. An overall increase of 8.78% in effectiveness was observed when both Human and Yeast PPIs were employed. The upsurge in the efficiency of complex prediction is measured using evaluation criteria such as geometric accuracy, sensitivity, precision, recall, and positive predicted values. The methods tend to exhibit accuracy of 0.213, 0.341, 0.220, 0.262 and 0.228 for Human PPIs when both topological and biological features were used. When topological information is used alone, the above approaches showed accuracy of just 0.130, 0.047, 0.050, 0.065, and 0.114. Therefore, the methods exhibit a significant performance improvement in detecting protein complexes when combining biological and topological information.