Answers are grounded in this paper and cite their sources.
Transcriptional profiling techniques facilitate the identification of biomarkers: genes that provide insight into disease diagnosis and natural history. Biomarker discovery can be modelled as a feature selection problem that aims to find the most discriminating features (i.e. gene expression patterns) for accurate disease classification, particularly in identifying those individuals at risk of more aggressive disease. Typical feature selection algorithms identify individual genes, and the disease discrimination power of each gene is considered separately. We describe a gene selection method incorporating prior biological knowledge of genetic pathways to find groups of strongly-correlated genes that accurately discriminate complex as well as simple disease traits. The first step is to identify the most impacted biological processes in a genome-wide expression (microarray) dataset describing a particular pathological condition. A specified number of differentially-expressed genes from relevant pathways are then selected for disease classification. The advantage of this method is that it searches for a related group of strongly-correlated genes rather than individual ones, with greater relevance to the disease type or clinical grade. We compared our algorithm to existing feature selection and ranking methods using open-access datasets derived from a series of well characterised clinical conditions, and two classification methods: K-Nearest Neighbour (KNN) and Support Vector Machine (SVM). Our algorithm consistently outperformed other methods in terms of disease classification accuracy. Moreover, we were able to reduce the number of genes required to accurately discriminate disease states, critical in the creation of diagnostic tests.
No takes yet. Share an insight, caveat, or question.
Ibrahim et al. (2011) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: