Enhancing Automatic Construction of Gene Subnetworks by Integrating Multiple Sources of Information


Suwannaroj, Sujimarn and Niranjan, Mahesan (2008) Enhancing Automatic Construction of Gene Subnetworks by Integrating Multiple Sources of Information. Journal of Signal Processing Systems, 50, (3), 331-340.

Download

[img] PDF
Download (28Kb)

Description/Abstract

We present an approach to extracting information from textual documents of biological knowledge and demonstrate how cellular gene pathways may be inferred. Natural language processing techniques are used to represent title and abstract fields of publications to derive a gene similarity vectors which are subject to cluster analysis. Gene interactions are derived by parsing sentences in the abstracts to infer causal relationships. We show how high throughput transcriptome data may then be used to enhance the construction of gene pathways from information derived from text. Subnetworks constructed by integrating information automatically derived from literature with gene expression data is validated by comparing biological processes defined in the Gene Ontology 2(GO) database. We find that precision increases in $$58\%$$ of the clusters when enhanced in this manner while a decrease in precision is observed in a relatively small number of clusters. These results are compared to similar attempts at the same problem and appear to be better in terms of precision of network construction. We also show an example of a subnetwork found by this analysis that overlaps a known gene pathway in KEGG and MIPS databases.

Item Type: Article
ISSNs: 1939-8018
Related URLs:
Divisions: Faculty of Physical Sciences and Engineering > Electronics and Computer Science > Comms, Signal Processing & Control
ePrint ID: 266705
Date Deposited: 24 Sep 2008 07:41
Last Modified: 27 Mar 2014 20:12
Publisher: Springer
Further Information:Google Scholar
ISI Citation Count:1
URI: http://eprints.soton.ac.uk/id/eprint/266705

Actions (login required)

View Item View Item