Integration of gene expression and DNA methylation data across different experiments

Integrative analysis of multi-omic datasets has proven to be extremely valuable in cancer research and precision medicine. However, obtaining multi-modal data from the same samples is often difficult. Integrating multiple datasets of different omics remains a challenge, with only a few available algorithms developed to solve it. Here, we present INTEND (IntegratioN of Transcriptomic and Epige-Nomic Data), a novel algorithm for integrating gene expression and DNA methylation datasets covering disjoint sets of samples. To enable integration, INTEND learns a predictive model between the two omics by training on multi-omic data measured on the same set of samples. In comprehensive testing on 11 TCGA (The Cancer Genome Atlas) cancer datasets spanning 4329 patients, INTEND achieves significantly superior results compared with four state-of-the-art integration algorithms. We also demonstrate INTEND's ability to uncover connections between DNA methylation and the regulation of gene expression in the joint analysis of two lung adenocarcinoma single-omic datasets from different sources. INTEND's data-driven approach makes it a valuable multi-omic data integration tool. The code for INTEND is available at https://github.com/Shamir-Lab/INTEND

Automated summary

The integration of multi-omic datasets is valuable in cancer research and precision medicine, but obtaining multi-modal data from the same samples is challenging. INTEND is a novel algorithm that integrates gene expression and DNA methylation datasets by learning a predictive model between the two omics. It achieves superior results compared to other integration algorithms and can uncover connections between DNA methylation and gene expression regulation.