Optimized-ANN for Predicting Cancer Using Gene Expression RNA-Seq Data

Early detection of cancer is important because it can greatly enhance the likelihood of effective treatment and cure. Researchers have developed numerous approaches based on gene expression for cancer detection. Gene expression is the method by which Deoxyribose Nucleic Acid (DNA) is transformed to Ribose Nucleic Acid (RNA), and then to protein. Several strategies are available in the literature for the classification of cancerous and noncancerous genes from gene expression data. The current study investigates the performance of Artificial Neural Networks (ANNs) to predict the cancer type on a publicly available UC Irvine Machine Learning Repository. This study provides a comparative analysis of different extensions of Adam optimizers like Adaptive Moment (Adam), Adaptive Gradient (AdaGrad), as well as Nesterov-accelerated Adaptive Moment Estimation (Nadam) and proposes the significance of optimized ANNs in the form of a computationally efficient model. The study also explores the impact of multiple learning and decay rates on the prediction performance of ANN.