Shortest path analysis using partial correlations for classifying gene functions from gene expression data

Abstract

Motivation: Gaussian graphical models (GGMs) are a popular tool for representing gene association structures. We propose using estimated partial correlations from these models to attach lengths to the edges of the GGM, where the length of an edge is inversely related to the partial correlation between the gene pair. Graphical lasso is used to fit the GGMs and obtain partial correlations. The shortest paths between pairs of genes are found. Where terminal genes have the same biological function intermediate genes on the path are classified as having the same function. We validate the method using genes of known function using the Rosetta Compendium of yeast (Saccharomyces Cerevisiae) gene expression profiles. We also compare our results with those obtained using a graph constructed using correlations.

Analyzing networks of phenotypes in complex diseases: methodology and applications in COPD

“…The most highly connected nodes were FEV1 and Gas Trapping (see Figure 1), with Gas Trapping significantly connected with all of the analyzed phenotypes. In addition, the 16 pairs that were not directly connected (p-values >0.001) were connected through only one transitive node based on shortest path analysis [18]. The majority of shortest paths connected through gas trapping (9 out of 16), suggesting that gas trapping is a “hub” in the phenotypic network.…”

A survey of the all-pairs shortest paths problem and its variants in graphs

“…The problem of finding the shortest distance between two vertices of a graph and finding a path that causes it are classic problems in graph algorithms. It appears in countless practical applications and is an important concept in transportation (and communication) engineering, computer science, network routing, network analysis [63], image processing [37,61], operation research [19, pages 657], VLSI design [66], DNA analysis [70], bio-informatics [39], chemical compounds [6,31], computational geometry and robotics [33], to mention few central areas of interest. Because of its rich in applications, the work on such problems are deep and vast (in all kinds of classic graphs, directed or undirected, weighted or unweighted), in both the scientific community and engineering community.…”

Breaking Free from the Chains of Pathway Annotation: de Novo Pathway Discovery for the Analysis of Disease Processes

“…It has been shown that genes or proteins with high network proximity tend to have similar biological functions [52] and several methods have been developed to predict gene function exploiting the colocalization of functionally annotated genes with genes of unknown function [53][54][55][56]. Network proximity can also be used as a predictor for genes implicated in disease processes [57,58].…”