Exploiting Indirect Neighbours and Topological Weight to Predict Protein Function from Protein-Protein Interactions

Abstract

Most approaches in predicting protein function from proteinprotein interaction data utilize the observation that a protein often share functions with proteins that interacts with it (its level-1 neighbours). However, proteins that interact with the same proteins (i.e. level-2 neighbours) may also have a greater likelihood of sharing similar physical or biochemical characteristics. We speculate that two separate forms of functional association accounts for such a phenomenon, and a protein is likely to share functions with its level-1 and/or level-2 neighbours. We are interested to find out how significant is functional association between level-2 neighbours and how they can be exploited for protein function prediction. We made a statistical study on recent interaction data and observed that functional association between level-2 neighbours is clearly observable. A substantial number of proteins are observed to share functions with level-2 neighbours but not with level-1 neighbours. We develop an algorithm that predicts the functions of a protein in two steps: (1) assign a weight to each of its level-1 and level-2 neighbours by estimating its functional similarity with the protein using the local topology of the interaction network as well as the reliability of experimental sources; (2) scoring each function based on its weighted frequency in these neighbours. Using leave-one-out cross validation, we compare the performance of our method against that of several other existing approaches and show that our method performs well.

A neural network algorithm for semi-supervised node label learning from unbalanced data

“…Algorithms based on the guilt-by-association principle set unlabeled nodes according to the majority of the labels in their direct neighborhoods (Marcotte et al, 1999;Oliver, 2000). By extending this approach, indirect neighbours, that account for pairs of nodes connected through intermediate ones, have been used to extend the notion of pairwise-similarities among nodes (Chua et al, 2006;Li et al, 2010;Bogdanov & Singh, 2010). Other methods focused on clustering nodes into functional modules based on the graph topology, and assigning to unlabeled nodes the most common labels in a given module (Sharan et al, 2007;Zhu et al, 2010).…”

Protein Interactome and Its Application to Protein Function Prediction

“…After that, the Gibbs sampling technique is iteratively applied to determine the stable values of this probability for each protein. This approach resulted in higher performance than those of neighbourhood-based approaches (Chua et al, 2006;Hishigaki et al, 2001;Schwikowski et al, 2000) when utilized to the yeast PPI data.…”

“…In another approach, the shared neighbourhood of a pair of proteins are considered besides from the neighbourhood of the protein of interest. Chua et al investigated the correlation between functional similarity and network distance (Chua et al, 2006). They developed a functional similarity score, called the FS-weight measure, which gives different weights to proteins depending on their network distance from the query protein.…”

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

“…ss (u, v) of each edge e (u, v) is normalized as follows: (2) B. A novel functional similarity based on topology and GO annotation Several approaches to PPI graph weighting use the graphtheoretic methods, such as CD-distance [3], FSWeight [19], and AdjustCD [3]. These methods were proposed based on the principle that the higher the number of common interactors shared by two proteins, the more likely they are functionally related.…”

“…where u and v are used to penalize proteins with very few neighbors as in FSWeight [19]. However, these approaches only use the topology of the graph to induce weighting.…”