Classifying protein complexes from candidate subgraphs using fuzzy machine learning model

Many computational methods have been applied to identify protein complexes from experimentally obtained protein-protein interaction (PPI) networks. Because of the presence of unreliable interactions in PPI networks, multi-functionality of proteins, and complex connectivity of the PPI network, the task is very challenging. In this study, we tackle the presence of unreliable interactions in protein complex using Genetic-Algorithm Fuzzy Naïve Bayes (GAFNB) which takes unreliability into consideration. Many existing methods can provide lots of candidate subgraphs. So we focused on how to classify the protein complexes from the subgraphs by considering the fuzzy attribute of PPI. We experimented with two datasets of size 10,371 and 986, each containing 493 positive protein complexes from MIPS and TAP-MS datasets. We compared the performance of GAFNB with Naïve Bayes (NB). Results show that GAFNB performed better which indicates that a fuzzy model is more suitable when unreliability is present. It is necessary to consider the unreliability in identifying protein complexes.