Identifying Significant Gene Interaction Networks Using Machine Learning and Statistical Techniques

Nowadays, a massive library of Data Mining techniques has been expanded to undertake a plenty of problems in fields such as medical imaging, traffic analysis, business administration, marketing and sales, manufacturing process astronomy and etc. At present, Data Mining had a major impact on the information industry, due to the broad availability of vast datasets. Classification technique is one category of broadly applied methods of data mining in healthcare. It utilized to classifies each item in a collection of data into one of a predefined collection of groups. The proposed research work focuses on categorizing the dengue microarray dataset using SVM classifier to identify the significant genes, discrimination between various classes present in the dengue dataset and interaction among the significant genes with constructing gene networks using Pearson Correlation Method. The Significant genes have been identified depends on gene ranking with its posterior probability threshold value 0.95 and there are 554, 172, 271 and 762 significant genes are extracted for convalescent, dengue hemorrhagic fever, dengue fever and healthy control respectively. From the confusion matrix observed that SVM Classification algorithm gives 100% accuracy for top 30 genes in each class of samples. From the top 30 classified genes in three sample groups, it is observed that 16 significant genes interacted with each other in Convalescent group, 18 significant genes interacted with each other in Dengue Hemorrhagic Fever group, 17 significant genes interacted with each other in Dengue Fever group and 20 significant genes interacted with each other in Healthy Control group.