Deep Learning for Breast Cancer Prediction in the Era of Big Data: A Comparative Study of Gene Expression and DNA Methylation
I. Jaichitra, T A Mohanaprakash, C Poonguzhali, S. Janagiraman, S Selvakumaran, B. Uma Maheswari · 2023
In the era of big data, conventional machine learning algorithms struggle to handle the vast and unique characteristics of this data-rich landscape. This study is dedicated to the challenging task of breast cancer prediction, focusing on two critical data types: gene expression (GE) and DNA methylation (DM). Our primary goal is to leverage the power of Deep Learning algorithms, both individually for each dataset and in combination,. At the heart of breast cancer prediction, System employ the Convolutional Neural Network (CNN) algorithm, renowned for its prowess in pattern recognition and feature extraction. Our research involves a comprehensive analysis of GE, DM, and the merged GE and DM datasets. Also evaluate the performance of the CNN algorithm by comparing it with the widely-used Random Forest (RF) algorithm for classification tasks. Experiments underscore the effectiveness of our approach, particularly when using the GE dataset. This configuration yields the highest accuracy and is the most cost-effective among all classifiers considered. These findings highlight the remarkable potential of Deep Learning techniques in breast cancer prediction, especially in the context of big data. This research aims to develop a deep learning framework for identifying significant breast cancer biomarkers. This model excels in managing large volumes of data, even when dealing with missing values and class imbalances. The study emphasizes the advantages of feature reduction and class balance, resulting in improved predictive outcomes compared to imbalanced datasets. Furthermore, conduct Gene Set Enrichment Analysis on the gene sets identified by the model and assess its efficacy by comparing these gene sets with established cancer resources. The results suggest that deep learning methods can be extended to predict various cancer types.