CWAOMT: Class Weight balanced Artificial Neural Network model for the Classification of Ovarian Malignancy from Transcriptomic Profiles

Abstract

The ability to accurately diagnose cancer is crucial to saving lives. Epithelial Ovarian Cancer (EOC) is a hard and serious disease that affects many women in worldwide. It has a poor prognosis and a molecular pathogenesis that is still unknown. Nowadays, RNA-Seq-based gene expression data have paved the way for more effective treatment in order to increase the early diagnosis of cancer. In this paper, a classweight balancing ANN is employed to detect recurrent ovarian cancer for RNA-Seq data. The model performed admirably, accurately classifying both primary and recurrent tumors without bias with 98% of accuracy rate. Later the DL model is saved using Python’s Pickle tool to avoid re-training and the pre-trained model generated for the same output. The proposed pretrained CWAOMT produced output within 12milliseconds as compared with 466milliseconds before pretraining. The experiment shows that the suggested CWAOMT performed better than the classification without data balancing. This pretrained model can be employed for later classifications of similar data without losing the achieved trained outcome.