Original title: Contextualized Networks Reveal Heterogeneous Transcriptomic Regulation in Tumors at Sample-Specific Resolution
Authors: Caleb N Ellington,Benjamin J Lengerich,Thomas BK Watkins,Jiekun Yang,Hanxi Xiao,Manolis Kellis,Eric P Xing
This article focuses on understanding the complexity of gene expression and regulation in cancer. Cancers are influenced by multiple factors, including genetic mutations, the surrounding microenvironment, and the patient’s background. These factors contribute to the heterogeneity of cancer cells and their behavior.
To better understand this regulation-driven heterogeneity, the authors propose a method called contextualized learning. This approach allows for the inference of sample-specific models by incorporating phenotypic, molecular, and environmental information into the gene regulatory network (GRN) models. The authors applied this approach to analyze data from 7997 tumors across 25 different types of cancer.
The contextualized GRN models provide a detailed and personalized view of gene expression dynamics, revealing co-expression modules and regulatory elements specific to each sample. This approach also allows for the prediction of GRNs for new tumor types based on a pan-cancer model.
Additionally, the authors demonstrate how these contextualized networks can be applied to precision oncology, uncovering the effects of gene networks on known biomarkers and improving survival prognosis for thyroid, brain, and gastrointestinal tumors.
It is important to note that the authors have declared no competing interests.
Original article: https://www.biorxiv.org/content/10.1101/2023.12.01.569658v1