Integration of gene expression and DNA methylation data across different experiments.

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Multi-omic data integration is crucial for cancer research and precision medicine.
• Acquiring multimodal data from identical samples is challenging.
• Existing algorithms for integrating disjoint omics datasets are limited.

Purpose of the Study:

• To introduce INTEND (IntegratioN of Transcriptomic and EpigeNomic Data), a novel algorithm for integrating gene expression and DNA methylation data from disjoint sample sets.
• To develop a method that learns predictive models between omics layers using shared samples.

Main Methods:

• INTEND trains a predictive model by learning associations between gene expression and DNA methylation on samples with both data types.
• The algorithm is tested on 11 The Cancer Genome Atlas (TCGA) datasets, comprising 4329 cancer patients.
• Performance is compared against four state-of-the-art integration algorithms.

Main Results:

• INTEND demonstrated significantly superior performance compared to existing integration methods across 11 TCGA cancer datasets.
• The study successfully uncovered novel connections between DNA methylation and gene expression regulation.
• INTEND facilitated the joint analysis of lung adenocarcinoma datasets from disparate sources.

Conclusions:

• INTEND is an effective and valuable tool for multi-omic data integration, particularly for disjoint datasets.
• The algorithm advances the ability to study gene expression regulation through epigenetic modifications.
• INTEND offers a data-driven approach to enhance insights in cancer genomics and precision medicine.