Artificial Intelligence-based Multiomics Integration Model for Cancer Subtyping
Aadil Rashid Bhat, Rana Hashmy · 2022
One out of every six deaths in the world is attributed to Cancer. On the whole, Cancer is a personalized and heterogeneous condition with a wide variety of subtypes, each with its characteristics and pathogeneses. Cancers differ in their clinical characteristics and survival times due to the distinct nature of similar tumors. Which renders them difficult to classify, predict and diagnose. Recently, joint analysis of cancers across multiple omics layers like genomics, epigenomics, and transcriptomics has provided a new perspective on cancer deregulation. Combined analysis of these complementary signals from different cellular functional layers has the potential to help in fine-tuned characterization, classification, and early diagnosis of cancers. However, in addition to the multitude of heterogeneous variables in omics data, multi-omics data pose many challenges, because of their disparate nature, in data integration and knowledge discovery. This study examines the challenge of multi-omics data integration using novel Deep learning (DL) methods like Deep Auto-encoder. We explore the application of Auto-encoders for multi-omics analysis to identify homogeneous cancer subtypes, sharing similar etiology, responses, and clinical outcomes.