Abstract
With the explosion of high-throughput data, effective integrative analyses are needed to decipher the knowledge accumulated in biological databases. Existing meta-analysis approaches in systems biology often focus on hypothesis testing and neglect real expression changes, i.e. effect sizes, across independent studies. In addition, most integrative tools completely ignore the topological order of gene regulatory networks that hold key characteristics in understanding biological processes. Here we introduce a novel meta-analysis framework, Network-Based Integrative Analysis (NBIA), that transforms the challenging meta-analysis problem into a set of standard pathway analysis problems that have been solved efficiently. NBIA utilizes techniques from classical and modern meta-analysis, as well as a network-based analysis, in order to identify patterns of genes and networks that are consistently impacted across multiple studies. We assess the performance of NBIA by comparing it with nine meta-analysis approaches: Impact Analysis, GSA, and GSEA combined with classical meta-analysis methods (Fisher’s and the additive method), plus the three MetaPath approaches that employ multiple datasets. The 10 approaches have been tested on 1,737 samples from 27 expression datasets related to Alzheimer’s disease, acute myeloid leukemia (AML), and influenza. For all of the three diseases, NBIA consistently identifies biological pathways relevant to the underlying diseases while the other 9 methods fail to capture the key phenomena. The identified AML signature is also validated on a completely independent cohort of 167 AML patients. In this independent cohort, the proposed signature identifies two groups of patients that have significantly different survival profiles (Cox p-value 2 × 10−6). The NBIA framework will be included in the next release of BLMA Bioconductor package (http://bioconductor.org/packages/release/bioc/html/BLMA.html).
Highlights
With the explosion of high-throughput data, effective integrative analyses are needed to decipher the knowledge accumulated in biological databases
We present a novel network-based meta-analysis that is able to combine multiple studies and identify the signaling pathways that are significantly impacted in a given phenotype
The main innovation of Network-Based Integrative Analysis (NBIA) is that it transforms the challenging meta-analysis problem into a set of standard analysis problems that can be solved efficiently. This approach utilizes techniques from both p-value-based and effect-size-based meta-analysis techniques in order to reliably identify a robust set of impacted genes. This set of genes serves as the input of the impact analysis (IA) approach to identify the biological processes that are significantly impacted under the effect of the disease
Summary
With the explosion of high-throughput data, effective integrative analyses are needed to decipher the knowledge accumulated in biological databases. Regardless of the high-throughput platforms being used, a standard comparative analysis of expression data usually produces a set of differentially expressed (DE) genes, which are often regarded as potential biological markers These genes are important in classifying and subtyping patients, as well as in identifying entities that may involve in biological processes of the underlying diseases[3,4,5,6]. We propose Network-Based Integrative Analysis (NBIA), a network-based approach that utilizes techniques from both p-values-based and effect-sizes-based methods to reliably identify genes and pathways that are likely to be impacted by the underlying disease. The estimated genome-scale expression change allows for topology-aware analysis, in which gene interaction and signal propagation are taken into consideration This approach transforms the meta-analysis problem into a standard topology-aware pathway analysis problem that has been solved efficiently. NBIA outperforms existing approaches in identifying biological processes relevant to the disease
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
