Assessment of prediction accuracy of protein function from protein–protein interaction data

Abstract

Functional prediction of open reading frames coded in the genome is one of the most important tasks in yeast genomics. Among a number of large-scale experiments for assigning certain functional classes to proteins, experiments determining protein-protein interaction are especially important because interacting proteins usually have the same function. Thus, it seems possible to predict the function of a protein when the function of its interacting partner is known. However, in vitro experiments often suffer from artifacts and a protein can often have multiple binding partners with different functions. We developed an objective prediction method that can systematically include the information of indirect interaction. Our method can predict the subcellular localization, the cellular role and the biochemical function of yeast proteins with accuracies of 72.7%, 63.6% and 52.7%, respectively. The prediction accuracy rises for proteins with more than three binding partners and thus we present the open prediction results for 16 such proteins.

Widely predicting specific protein functions based on protein-protein interaction data and gene expression profile

“…The comparison study shows that we can get better predicting performance based on filtered interaction network. There are some classical methods such as the neighboring counting method [8] and chi-square method [9] to predict functions by considering its neighboring interactions. When their functional annotations are complex, due to the absence of considering the relations among classes, employing distant neighbors in interaction network will decrease the prediction accuracy [16] .…”

“…Therefore, an important task for bioinformatics is to conduct proteome-scaled functional assignments for a large number of unknown proteins. The traditional approaches to predicting uncharacterized proteins are mainly based on sequence homology analysis [2,3] , while many recently developed approaches are able to exploit high-throughput data for gene expression [4][5][6][7] and protein-protein interaction [8][9][10][11][12] for functional predictions. Some of the approaches attempt to integrate different kinds of data to construct function-related networks [13][14][15][16] for helping predict proteins.…”

Protein function prediction using guilty by association from interaction networks

“…For example, the χ 2 test has been used to rank the functional terms associated to a group of interacting partners by comparing the frequency of the terms within the group and with the expected distribution in the whole network (Hishigaki et al 2001). Another work, PRODISTIN (Brun et al 2003), focuses on the clusterization of the entire Yeast interaction graph by means of a distance measure to define groups associated with the same functional class.…”

“…Approaches exploiting interaction networks have been widely used for annotation of the Yeast genome (Hishigaki et al 2001;Brun et al 2003;Deng et al 2003;Nabieva et al 2005;Chua et al 2006). At the same time, many tools which analyze biological network properties are already available.…”

On network-based kernel methods for protein-protein interactions with applications in protein functions prediction

“…The method predicts for a given protein up to three functions that are most common among its neighbors. The Chi-square method [10] , it computes the Chi-square scores of function assignment and assign the functions with several largest scores to a given protein. Vazquez, et al [11] , Karaoz [12] and Nabieva [13] applied graph algorithms such as cut-based approach and flow-based approach for functional analysis.…”