Posts Tagged ‘mur0mg’

Reconciling differential gene expression data with molecular interaction networks

January 28, 2015

Reconciling differential gene expression
w/…#networkshttp://bioinformatics.oxfordjournals.org/content/29/5/622 Propagating this across interactions finds perturbed pathways

This paper basically propagates scores of disease-related highly differentially expressed genes (-log10 p) over human protein interaction network, calculates new scores using four major algorithms (Vanilla, PageRank, GeneMANIA, Heat Kernel), re-ranks genes based on the new scores and then finds enriched pathways among top-ranking genes. Compared with traditional ways by ranking highly differentially expressed genes based on p-values without any network information, the approach not only recovered canonical pathways but also discovered novel ones such as an insulin-mediated glucose transport pathway in Huntington’s disease. The authors also explored differences among four algorithms and identified the top-ranking genes specifically found by particular algorithms. In short, the paper provides a valuable framework for integrating networks and gene expression data. Their analysis for comparing four major algorithms is also helpful.

http://bioinformatics.oxfordjournals.org/content/29/5/622

VIRGO: computational prediction of gene functions

December 25, 2014

VIRGO: computational prediction of gene function http://nar.oxfordjournals.org/content/34/suppl_2/W340.full Webserver propagates GO terms over PPI & gene-expression #networks

This work was said to be the first web server for gene function
annotation (not the first algorithm).
The idea is to predict gene functions from known molecular interaction
networks (such as PPI), which includes both annotated and unannotated
genes. The potential function of an unannotated gene is predicted
using a propagation diagram, which takes into account the neighbors’
functions. The weight of edge in the network is determined by user uploaded expression data. Weight = |Pearson correlation| of expression
profiles of the gene pair. Weight reflects the confidence of the edge.

BMC Bioinformatics | Abstract | Sensitive detection of pathway perturbations in cancers

December 23, 2014

Sensitive detection of pathway perturbations in #cancers
http://www.biomedcentral.com/1471-2105/13/S3/S9/abstract Differential expression of #pathways (in toto or a sub-part)

** Sensitive detection of pathway perturbations in cancers. Rivera et al. BMC Bioinfo (2014)

In this paper, the authors introduce a new computational method that identifies subsets of pathways that exhibit differential gene expression between cancer and normal tissue. Some previous methods only considered differential expression in sets of genes without considering the structure of interactions between the genes or their protein products. Other previous methods looked at pathway
perturbations, but required all members of a pathway to exhibit differential gene expression in order for the pathway to appear significant. The authors demonstrate the general superiority of their method to these previous methods, as well as the robustness of their method to missing data. The authors also consider future enhancements, such as taking into account the direction of differential expression, using more information on the nature of each gene interaction involved, and using universal protein interaction networks to incorporate data beyond what is found in curated pathway databases.

Network Legos: Building Blocks of Cellular Wiring Diagrams | Abstract

November 1, 2014

#Network Legos: Building Blocks of Cellular Wiring Diagrams
http://online.liebertpub.com/doi/abs/10.1089/cmb.2007.0139 Active subnets from protein interactions & co-expression

This paper describes a top-down, set-theoretic approach to comparing gene expression network dynamics under multiple conditions. The method takes an input “wiring diagram” of gene interactions as well as gene expression datasets, which are used to derive interaction profiles composed of enriched genes. Sets of these interaction profiles are used to identify statistically significant “network blocks”
representing network modules underlying the various enriched profiles. These blocks are then organized in a directed acyclic graph in order to identify “network legos”, which represent the most fundamental building blocks of the enriched gene expression networks. The authors demonstrate the utility of the method by identifying a differentially enriched pathway in various forms of leukemia and finding common network modules activated by two human cell types in response to different cellular stresses.

http://online.liebertpub.com/doi/abs/10.1089/cmb.2007.0139

Signaling hypergraphs: Trends in Biotechnology

October 9, 2014

Signaling #hypergraphs
http://www.cell.com/trends/biotechnology/abstract/S0167-7799(14)00071-7 Edges from interactions of 2 sets of nodes. Better representation of assemblies & #complexes.

QT:{{”
each edge is defined not by interaction of 2 nodes (as in graphs), but 2 sets of nodes (known as hypernodes in hypergraphs)……The use of hypernodes also represents three concepts better than directed or non-directed graphs: protein complexes, protein assemblies and regulation (especially involving complexes/assemblies).
“}}

Signaling hypergraphs. Ritz et al. (2014) TIB

This opinion paper advocates the use of hypergraphs to complement graph-based signaling network and pathway analyses, where each edge is defined not by interaction of 2 nodes (as in graphs), but 2 sets of nodes (known as hypernodes in hypergraphs). They argue that
hypergraphs is a set-based method that acts like a more general version of a graph. The use of hypernodes also represents three concepts better than directed or non-directed graphs: protein complexes, protein assemblies and regulation (especially involving complexes/assemblies). They propose that hypergraphs can be very useful in situations where the effects of individual proteins might be neglected in graphs but will have a noticeable effect when these proteins are included in protein complexes as hypernodes. They use 3 applications as examples: pathway enrichment, pathway reconstruction, and pathway crosstalk.