Integrative bioinformatics analyses of genome-wideRNAi screens

Abstract

In past few years, genome-wide RNAi screens have identified many novel genes involved in diseases for many viruses such as Human Immunodeficiency Virus-1 (HIV-1), Hepatitis C virus (HCV), West Nile Virus (WNV) and Influenza virus (IV). However, due to difference in experimental conditions, usage of different viral strains and inherent biological noise, these screens have shown low number of common or overlapping hits for a virus. Moreover, this overlap gets poorer for similar studies on viruses of different families. Although these overlaps are significant, their lower size restricts a comprehensive insight from a comparative analysis. Thus, a direct comparison of gene hit-lists of RNAi screens may not always give meaningful results. To address this problem we propose an integrative bioinformatics pipeline that allows for network based meta-analysis of viral HT-RNAi screens. Initially, human protein interaction network (PIN) generated by collating data from various public repositories, is subjected to unsupervised clustering to determine functional modules. Those modules that are significantly enriched in host dependency factors (HDFs) and/or host restriction factors (HRFs) are then filtered based on network topology and semantic similarity measures. Modules passing all these criteria are then interpreted for their biological significance from enrichment analyses. With our approach we could predict Tankyrase-1 as a potential novel hit within the functional subnetworks, within the human PIN for Hepatitis C virus (HCV). and Human Immunodeficiency Virus-1 (HIV-1), based on HDFs and HRFs identified in the corresponding genome-wide RNAi screens of these viruses. Thus, our approach allows for a network based meta-analysis of genome-wide screens to develop plausible hypotheses for novel regulatory mechanisms in virus-host interactions based on RNAi screens.