Deep Learning for Human Disease Detection, Subtype Classification, and Treatment Response Prediction Using Epigenomic...

Area of Science:

• Computational Biology
• Genomics
• Machine Learning

Background:

• Deep learning (DL) offers advanced capabilities for analyzing complex biological data.
• The application of DL in human epigenomics for disease prediction is an emerging field.
• Epigenomic data, including DNA methylation and RNA sequencing, holds significant potential for disease insights.

Purpose of the Study:

• To critically review studies utilizing DL models for disease detection, subtype classification, and treatment response prediction from epigenomic data.
• To assess the performance and identify trends in DL applications within translational epigenomics.
• To propose a workflow for developing predictive models in this domain.

Main Methods:

• Systematic literature search across major scientific databases (PubMed, Scopus, Web of Science, Google Scholar, arXiv.org).
• Adherence to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.
• Inclusion of 22 relevant studies out of 1140 identified publications.

Main Results:

• Deep learning models demonstrated high predictive accuracy: 88.3%-100.0% for disease detection, 69.5%-97.8% for subtype classification, and 80.0%-93.0% for treatment response.
• DNA methylation and RNA-sequencing data were the most commonly used epigenomic data types.
• A comprehensive workflow for developing predictive models was generated, covering all stages from task definition to performance evaluation.

Conclusions:

• Deep learning is a powerful tool for extracting valuable knowledge from epigenomic big data.
• DL applications in epigenomics show significant promise for enhancing disease prediction and personalized medicine.
• Further research and development in DL for translational epigenomics are warranted to fully realize its potential.