Email updates

Keep up to date with the latest news and content from BMC Genomics and BioMed Central.

Open Access Research article

Gene prioritization in Type 2 Diabetes using domain interactions and network analysis

Amitabh Sharma1, Sreenivas Chavali1, Rubina Tabassum1, Nikhil Tandon2 and Dwaipayan Bharadwaj1*

  • * Corresponding author: Dwaipayan Bharadwaj db@igib.res.in

  • † Equal contributors

Author Affiliations

1 Functional Genomics Unit, Institute of Genomics and Integrative Biology, CSIR, Delhi, India

2 Department of Endocrinology, All India Institute of Medical Sciences, New Delhi, India

For all author emails, please log on.

BMC Genomics 2010, 11:84  doi:10.1186/1471-2164-11-84

Published: 2 February 2010

Abstract

Background

Identification of disease genes for Type 2 Diabetes (T2D) by traditional methods has yielded limited success. Based on our previous observation that T2D may result from disturbed protein-protein interactions affected through disrupting modular domain interactions, here we have designed an approach to rank the candidates in the T2D linked genomic regions as plausible disease genes.

Results

Our approach integrates Weight value (Wv) method followed by prioritization using clustering coefficients derived from domain interaction network. Wv for each candidate is calculated based on the assumption that disease genes might be functionally related, mainly facilitated by interactions among domains of the interacting proteins. The benchmarking using a test dataset comprising of both known T2D genes and non-T2D genes revealed that Wv method had a sensitivity and specificity of 0.74 and 0.96 respectively with 9 fold enrichment. The candidate genes having a Wv > 0.5 were called High Weight Elements (HWEs). Further, we ranked HWEs by using the network property-the clustering coefficient (Ci). Each HWE with a Ci < 0.015 was prioritized as plausible disease candidates (HWEc) as previous studies indicate that disease genes tend to avoid dense clustering (with an average Ci of 0.015). This method further prioritized the identified disease genes with a sensitivity of 0.32 and a specificity of 0.98 and enriched the candidate list by 6.8 fold. Thus, from the dataset of 4052 positional candidates the method ranked 435 to be most likely disease candidates. The gene ontology sharing for the candidates showed higher representation of metabolic and signaling processes. The approach also captured genes with unknown functions which were characterized by network motif analysis.

Conclusions

Prioritization of positional candidates is essential for cost-effective and an expedited discovery of disease genes. Here, we demonstrate a novel approach for disease candidate prioritization from numerous loci linked to T2D.