Abstract:Inference of the gene regulatory networks arises as a fundamental problem in biomedical research. It is necessary to address this problem in order to explore the complex network of biomolecule interactions and regulations in living cells. Correspondingly, a large number of computational and statistical approaches, e.g. Bayesian Networks [1], Gaussian Graphical models [2]–[3], Relevance Networks [4] etc., have been proposed towards the reverse engineering of the underlying directed and/or undirected network topology. These approaches typically rely on experimental measurements from microarray and other high-throughput platforms which are often expensive and/or unable to capture complete information underneath. In this paper, we considered the problem of reconstructing directed networks using symbol data comprised of gene sets. We hypothesized that a large network is composed of several overlapping signaling transduction events with a linear arrangement of genes in each event. Gene sets in symbol data are defined by the group of genes participating in signaling transduction events, with the order of genes randomly permuted within each set (Fig. 1). The underlying network can be reconstructed by inferring the order of genes in each gene set. With few exceptions [5][6], the existing algorithms to network reconstruction do not accommodate symbol data in the form of unordered gene sets. We proposed a computational approach developed under the Gibbs sampling framework which leads to network reconstruction by treating unordered gene sets as random variables and sampling networks from their joint distribution. We assessed the robustness of our approach by performing sensitivity analyses on gold standard networks available from DREAM initiative (http://wiki.c2b2.columbia.edu/dream). We also observed the superior performance of the Gibbs sampling approach in an exhaustive simulation study in terms of sensitivity, specificity and PPV(positive predictive value), where we compared the performance of the proposed approach with two bayesian network approaches K2 [7] and MCMC [8]. Finally, we used our algorithm to reconstruct network in breast cancer cells.

Reciprocal Graphical Models for Integrative Gene Regulatory Network Analysis

Bayesian graphical models for computational network biology

Heterogeneous reciprocal graphical models

Application of new probabilistic graphical models in the genetic regulatory networks studies

Reverse Engineering Yeast Gene Regulatory Networks using Graphical Models

A Log-Linear Graphical Model for Inferring Genetic Networks from High-Throughput Sequencing Data

NJGCG: A node-based joint Gaussian copula graphical model for gene networks inference across multiple states

Assisted Graphical Model For Gene Expression Data Analysis

Integration of Epigenetic Data in Bayesian Network Modeling of Gene Regulatory Network

Integrating Prior Knowledge with Graph Encoder for Gene Regulatory Inference from Single-cell RNA-Seq Data

Integrative Analysis of Genetical Genomics Data Incorporating Network Structures.

Gene Regulatory Network Reconstruction Using Conditional Mutual Information

Assisted estimation of gene expression graphical models.

A Gaussian Graphical Model for Identifying Significantly Responsive Regulatory Networks from Time Series Gene Expression Data

Recovering Genetic Regulatory Networks by Integrating Multiple Data Sources

Gaussian Graphical Model for Identifying Significantly Responsive Regulatory Networks from Time Course High-Throughput Data

A Joint Graphical Model for Inferring Gene Networks Across Multiple Subpopulations and Data Types

Practical Guidelines for Incorporating Knowledge-Based and Data-Driven Strategies into the Inference of Gene Regulatory Networks

High-Dimensional Ordinary Differential Equation Models for Reconstructing Genome-Wide Dynamic Regulatory Networks

A Novel Computational Framework to Reconstruct Gene Regulatory Networks from Gene Sets

A sparse conditional Gaussian graphical model for analysis of genetical genomics data