A new essential protein discovery method based on the integration of protein-protein interaction and gene expression data

Abstract

BackgroundIdentification of essential proteins is always a challenging task since it requires experimental approaches that are time-consuming and laborious. With the advances in high throughput technologies, a large number of protein-protein interactions are available, which have produced unprecedented opportunities for detecting proteins' essentialities from the network level. There have been a series of computational approaches proposed for predicting essential proteins based on network topologies. However, the network topology-based centrality measures are very sensitive to the robustness of network. Therefore, a new robust essential protein discovery method would be of great value.ResultsIn this paper, we propose a new centrality measure, named PeC, based on the integration of protein-protein interaction and gene expression data. The performance of PeC is validated based on the protein-protein interaction network of Saccharomyces cerevisiae. The experimental results show that the predicted precision of PeC clearly exceeds that of the other fifteen previously proposed centrality measures: Degree Centrality (DC), Betweenness Centrality (BC), Closeness Centrality (CC), Subgraph Centrality (SC), Eigenvector Centrality (EC), Information Centrality (IC), Bottle Neck (BN), Density of Maximum Neighborhood Component (DMNC), Local Average Connectivity-based method (LAC), Sum of ECC (SoECC), Range-Limited Centrality (RL), L-index (LI), Leader Rank (LR), Normalized α-Centrality (NC), and Moduland-Centrality (MC). Especially, the improvement of PeC over the classic centrality measures (BC, CC, SC, EC, and BN) is more than 50% when predicting no more than 500 proteins.ConclusionsWe demonstrate that the integration of protein-protein interaction network and gene expression data can help improve the precision of predicting essential proteins. The new centrality measure, PeC, is an effective essential protein discovery method.

Consolidation and translation regulation: Figure 1.

“…Furthermore, exposing cortical neurons to glutamate, NMDA, or AMPA, or hippocampal slices to NMDA results in a rapid increase in p-eEF2 (Marin et al 1997;Belelovsky et al 2005), and blocking NMDARs results in a decrease, in conjunction with rapid translation of BDNF and Arc (Autry et al 2011), implying a role for phosphorylation of eEF2 in mediating glutamate receptor-mediated excitotoxicity (Hardingham and Bading 2010). In addition, blockade of NMDA receptor by the nonselective antagonist ketamine in the PFC has antidepressant effects and leads to increased synaptic protein synthesis via activation of the mTOR/S6K pathway and 4E-BP1, and prevents spine atrophy (Duman et al 2012;Li et al 2012). Furthermore, NMDA stimulation of hippocampal slices can also increase eIF4E phosphorylation in the CA1 region (Banko et al 2004).…”

“…Therefore, efforts are being made to create principled and biologically meaningful representations of these large-scale data in models that are flexible enough to Systems Biology modeling has been widely used in biology for many years; it frequently comprises just a single data type (for example, mRNA level or protein concentration) or uses small numbers of molecules or canonical pathways and rarely takes spatial constrains into consideration. More recently, integrative methods have begun to overlay multiple data sources onto these models, for example, visualizing mRNA expression data in the context of protein-interaction networks (Alcaraz et al 2012;Li et al 2012) or proteomic data (Hallock and Thomas 2012), but these methods of data integration do not implicitly model the relationships between the different data types, and the functional insight obtained is limited.…”

Identification of key biomarkers involved in osteosarcoma using altered modules

“…Gene interactions in the PPI sub-networks (occurring under OS and normal conditions) were re-weighted using PCC (value ranging from -1 to +1) (Li et al, 2012). Additionally, we defined the absolute value of PCC of each gene-gene interaction as the value of the interaction in the PPI sub-network.…”

A network approach predicts NFKBIA and BIRC3 as pathogenic genes in childhood asthma

“…The higher the weight score of a gene, the more it is involved in disease pathogenesis. Co-expression was evaluated by the Pearson correlation coefficient, which was a measure of the correlation between our predicted pathogenic gene and seed genes; this value ranged from -1 to +1 (Li et al, 2012). The weight w(x) of each gene x was computed as follows:…”