Feature prioritisation on big genomic data for analysing gene-gene interactions. (19th April 2021)

• Record Type: Journal Article
• Authors: Aloqaily, Ahmad A.; Tafavogh, Siamak; Harvey, Bronwyn L.; Catchpoole, Daniel R.; Kennedy, Paul J.
• Is Part Of: International journal of bioinformatics research and applications. Volume 17:Number 2(2021)
• Page Start: 158
• Page End: 177
• Publication Date: 2021-04-19
• Subjects: large genomic data -- dimensionality reduction -- feature prioritisation -- gene-gene interaction; Bioinformatics -- Periodicals; 570.285
• Languages: English
• ISSNs: 1744-5485

Abstract

Complex diseases are not caused by single genes but result from intricate non-linear interactions among them. There is a critical need to implement approaches that take into account non-linear gene-gene interactions in searching for markers that jointly cause diseases. Determining the interaction between more than two single nucleotide polymorphisms (SNP) within the whole genome data is computationally expensive and often infeasible. In this paper, we develop an approach to classify patients with Acute Lymphoblastic Leukaemia by analysing multiple SNP interactions. A novel feature prioritisation algorithm called interaction effect quantity (IEQ) selects SNPs with high potential of interaction by analysing their distribution throughout the genomic data and enables deeper analysis of non-linear interactions within large datasets. We show that IEQ enables analyses of interactions between up to four SNPs, with F-measure for classification greater than 89% obtained. Such an analysis is typically much more computationally challenging if IEQ is not implemented.