Deep learning challenges regarding genomics

The recent explosion of genomics data due to the advancements in next-generation sequencing (NGS) coupled with improvements in omic technologies (transcriptomics, proteomics, and metabolomics) has led to a greater understanding of the biological process of the living cell. Meanwhile, the remarkable success of DL based on DNN, has brought enormous improvements in computer vision (CV), natural language processing (NLP), and machine translation, and this has attracted the attention of genomics. The field of genomics quickly leveraged these specialized neural network architectures that can perform various tasks, such as binding site identification using CNNs, improving code optimization for improved protein translation through RNNs, unsupervised DL through autoencoders to predict gene expression, and so on. This is particularly exciting because genomics requires a data-driven and sophisticated solution to extract meaningful biological insights...