Read the original document by opening this link in a new tab.
Table of Contents
1 Introduction
2 Deep Learning Architectures: A Genomic Perspective
2.1 Convolutional Neural Networks
2.2 Recurrent Neural Networks
2.3 Autoencoders
2.4 Emergent Deep Architectures
2.4.1 Beyond Classic Models
2.4.2 Hybrid Architectures
2.5 Transformer-based Large Language Models
3 Deep Learning Architectures: Insights and Remarks
3.1 Model Interpretation
3.2 Transfer Learning and Multitask Learning
3.3 Multi-view Learning
4 Genomic Applications
4.1 Gene expression
4.1.1 Gene expression Characterization
4.1.2 Gene expression Prediction
4.2 Regulatory Genomics
4.2.1 Promoters and Enhancers
4.2.2 Splicing
4.2.3 Transcription Factors and RNA-binding Proteins
4.3 Functional Genomics
4.3.1 Mutations and Functional Activities
4.3.2 Subcellular Localization
4.4 Structural Genomics
4.4.1 Structural Classification of Proteins
4.4.2 Protein Secondary Structure
4.4.3 Contact Map
4.4.4 Protein Tertiary Structure and Quality Assessment
5 Challenges and Opportunities
5.1 The Nature of Data
5.1.1 Class-Imbalanced Data
5.1.2 Various Data Types
5.1.3 Heterogeneity and Confounding Correlations
5.2 Feature Extraction
5.2.1 Mathematical Feature Extraction
5.2.2 Feature Representation
6 Conclusion and Outlook
Summary
This paper provides a concise overview of deep learning applications in genomic research. It discusses the strengths of different deep learning models from a genomic perspective, fitting each particular task with the proper deep architecture. The paper also remarks on practical considerations of developing deep learning architectures for genomics. Additionally, it reviews deep learning applications in various aspects of genomic research, highlighting current challenges and potential research directions for future genomics applications.