Spiking neural networks for cancer gene expression time series modelling and analysis

Abstract

Gene expression can be used for profiling of cancer cell state and classification of disease. Some cancer variants have been attributed to one or few significant gene expression features. This paper investigates the combination of novel features selection methods - Minimum-Redundancy, Maximum-Relevance - and artificial neural networks - the spiking neural network NeuCube architecture - for genomic data classification and analysis. A NeuCube model performs not only a better classification than other machine learning methods, but most importantly contributes to the feature extraction and marker discovery along with providing gene interaction network analysis for selected genes. Results demonstrated that the methodology proposed could contribute to bioinformatics data analysis for the treatment of disease by discovery of new biomarkers from gene expression data.