Providing SSPCO Algorithm to Construct Static Protein-Protein Interaction (PPI) Networks

Abstract

Protein-Protein Inter-action Networks are dynamic in reality; i.e. Inter-actions among different proteins may be ineffective in different circumstances and times. One of the most crucial parameters in the conversion of a static network into a temporal graph is the well-tuning of transformation threshold. In this part of the article, using additional data, like gene expression data in different times and circumstances and well-known protein complexes, it is tried to determine an appropriate threshold. To accomplish this task, we transform the problem into an optimization one and then we solve it using a meta-heuristic algorithm, named Particle Swarm Optimization (SSPCO). One of the most important parts in our work is the determination of interestingness function in the SSPCO. It is defined as a function of standard complexes and gene co-expression data. After producing a threshold per each gene, in the following section we will discuss how using these thresholds, active proteins are determined and then temporal graph is created. For final assessment of the produced graph quality, we use graph clustering algorithms and protein complexes determination algorithms. For accomplishing this task, we use MCL, Cluster One, MCODE algorithms. Due to high number of the obtained clusters, the obtained results, if they have some special conditions, will filter out or be merged with each other. Standard performance criteria like Recal, Precision, and F-measure are employed. There is a new proposed criterion named Smoothness. Our experimental results show that the graphs produced by the proposed method outperform the previous methods.

Keywords

• p-p interaction networks
• dynamic networks
• sspco
• graph partitioning
• gene expression