BNrich: A Bayesian network approach to the pathway enrichment analysis

Abstract

Motivation: One of the most popular techniques in biological studies for analyzing high throughput data is pathway enrichment analysis (PEA). Many researchers apply the existing methods without considering the topology of pathways or at least they have overlooked a significant part of the structure, which may reduce the accuracy and generalizability of the results. Developing a new approach while considering gene expression data and topological features like causal relations regarding edge directions will help the investigators to achieve more accurate results. Results: We proposed a new pathway enrichment analysis based on Bayesian network (BNrich) as an approach in PEA. To this end, the cycles were eliminated in 187 KEGG human signaling pathways concerning intuitive biological rules and the Bayesian network structures were constructed. The constructed networks were simplified by the Least Absolute Shrinkage Selector Operator (LASSO), and their parameters were estimated using the gene expression data. We finally prioritize the impacted pathways by Fisher's Exact Test on significant parameters. Our method integrates both edge and node related parameters to enrich modules in the affected signaling pathway network. In order to evaluate the proposed method, consistency, discrimination, false positive rate and empirical P-value criteria were calculated, and the results are compared to well-known enrichment methods such as signaling pathway impact analysis (SPIA), bi-level meta-analysis (BLMA) and topology-based pathway enrichment analysis (TPEA). Availability: The R package is available on carn. Contact: kkavousi@ut.ac.ir

An integrative Bayesian network approach to highlight key drivers in systemic lupus erythematosus

“…According to BNrich properties, we could integrate two levels of biological data, gene expression and signaling pathways. Hence, the value of the node and edge parameters in addition to being influenced by gene expression data also depends on the topology of the underlying pathway [18]. For instance, in the Illumina platform, the mean value of the final parameter in PIK3CB gene was 1.7 and − 0.5, and the mean value of the final parameter in edge PPP3CC → NFATC3 was − 0.07 and 0.12, in TCR and BCR signaling pathways, correspondingly (Additional file 2).…”

“…In the parameter estimate step, the mean value of the expression for each gene (node) can be modeled as a linear regression of its parents' (upstream) gene expression [18]. When Y gene has X 1 , X 2 …X p − 1 parents in the Fig.…”

“…By applying gene expression data of SLE patients, Li et al recently reconstructed a BN, in which prior edges were randomly sampled based on the text-mining in SLE pathogenesis [12]. Remarkably, some methods such as BNrich were also developed based on BN properties [18] which identifies significant genes (nodes) and biological relationships (edges). The BNrich method has two key concepts: (1) the structure of BN is reconstructed by the signaling pathway structure, and (2) the expression level of the gene (node) is modeled as a regression function of expression levels of its upstream genes (parents) [18].…”

An integrative Bayesian network approach to highlight key drivers in systemic lupus erythematosus

“…The three desired pathways were exploited as structures of BNs [21]. In the parameter estimate step, the mean value of gene expression for each node can be modelled as a linear regression of its parents' gene expression because the gene expression data is continuous [22,24].…”

“…Remarkably, some pathway enrichment analysis (PEA) methods such as BNrich were also developed based on BN properties [19][20][21]. In addition, to detect significant pathways, BNrich also identifies significant genes (nodes) and biological relationships (edges).…”