International Journal of Pharma and Bio Sciences
ijpbs.net
editorijpbs@rediffmail.com (or) editorofijpbs@yahoo.com (or) prasmol@rediffmail.com
10.22376/ijpbs.2019.10.1.p1-12
Volume 7 Issue 3
2016 (July - September)
SNR-TR GENE RANKING METHOD: A SIGNAL-TO-NOISE RATIO BASED GENE SELECTION ALGORITHM USING TRACE RATIO FOR GENE EXPRESSION DATA
In this paper, we propose a hybridization of two popular gene selection methods, i.e. Trace Ratio (TR) algorithm and Signal-to-Noise Ratio (SNR) method. The proposed model defines two basic phases: in the first phase, the gene would be ranked using TR algorithm where the scoring weight matrix is re-defined using the SNR scoring technique (known as SNR-TR gene ranking method); and the second phase describes the validation of the top ranked genes using two variants of Neural Network (NN) classifier called resilient propagation and back propagation. It was observed that the SNR-TR gene ranking method found a subset of informative genes from the huge data of available genes and it is acknowledged by the maximum accuracy with less number of iterations obtained as compared to the existing TR algorithm. Further, our model have been experimentally analyzed using five benchmark datasets like Colon, Leukemia, Medulloblastoma, Lymphoma and Prostate Cancer. K number of top ranked genes were extracted using SNR-TR algorithm where K is defined as 50, 100 and 150. Though the technique is validated using different types of classifier, it can still be validated by using various performance index metrics like Balanced Classification Rate, Balanced Error Rate, Stability index etc.
SHRUTI MISHRA, DEBAHUTI MISHRA
Gene Regulatory Network, Gene Selection, Trace Ratio Algorithm, Signal-to-Noise Ratio, Classificatio
967-978