Using Indirect Protein-Protein Interactions for Protein Complex Prediction

Abstract

Protein complexes are fundamental for understanding principles of cellular organizations. Accurate and fast protein complex prediction from the PPI networks of increasing sizes can serve as a guide for biological experiments to discover novel protein complexes. However, protein complex prediction from PPI networks is a hard problem, especially in situations where the PPI network is noisy.We know from previous work that proteins that do not interact, but share interaction partners (level-2 neighbors) often share biological functions. The strength of functional association can be estimated using a topological weight, FS-Weight. Here we study the use of indirect interactions between level-2 neighbors (level-2 interactions) for protein complex prediction. All direct and indirect interactions are first weighted using topological weight (FS-Weight). Interactions with low weight are removed from the network, while level-2 interactions with high weight are introduced into the interaction network. Existing clustering algorithms can then be applied on this modified network. We also propose a novel algorithm that searches for cliques in the modified network, and merge cliques to form clusters using a "partial clique merging" method.In this paper, we show that 1) the use of indirect interactions and topological weight to augment proteinprotein interactions can be used to improve the precision of clusters predicted by various existing clustering algorithms; 2) our complex finding algorithm performs very well on interaction networks modified in this way. Since no any other information except the original PPI network is used, our approach would be very useful for protein complex prediction, especially for prediction of novel protein complexes.

Methods for protein complex prediction and their contributions towards understanding the organisation, function and dynamics of complexes

“…see [55]). Besides, the possibility of a predicted complex matching a reference complex that is small purely by chance is relatively high [61], and therefore evaluating the methods becomes challenging (further discussed in Section 2.4.3). Figure 1 shows the performance of nine methods using precision, recall, F-measure and AUC.…”

A Novel Protein Complex Identification Algorithm Based on the Integration of Local Network Topology and Gene Ontology

“…We compared our method with six previous competing algorithms, such as CMC [3], Core [10], PCP [6], IPCA [4], Coach [8], and CFinder [2] with recommended parameters. Furthermore, we evaluate the experimental result with four evaluation metrics, namely, specificity (Sp), sensitivity (Sn), f-score, and p-value, which are described in a previous study [20].…”

“…Protein complexes generally correspond to dense subgraphs in a PPI network because proteins in the same complex are highly interactive with one another. Many algorithms, such as Cfinder [2], CMC [3], IPCA [4], MCODE [5], and PCP [6], based on the hypothesis have been proposed to discover protein complexes.…”

“…Protein complexes are then detected on the basis of minimum cut in the weighted network. We compare our method with CMC [3], Core [10], PCP [6], Cfinder [2], Coach [8], and IPCA [4] in terms of specificity, sensitivity, and f-score by using two datasets of yeast protein interactions, Krogan [15] and DIP [16]. The experimental results show that our approach is more efficient than these algorithms.…”

Robust Community Detection Methods with Resolution Parameter for Complex Detection in Protein Protein Interaction Networks

“…This threshold is also used in other works, e.g. [20,4]. When the cluster and complex have the same size, a match then corresponds to the intersection containing at least half of the nodes in the complex and cluster.…”