A deep learning approach reveals unexplored landscape of viral expression in cancer

About 15% of human cancer cases are attributed to viral infections. Existing strategies to characterize viral expression in tumors largely rely on aligning RNA-seq reads to known viral databases, limiting the discovery of divergent viruses. The authors develop viRNAtap/viRNATrap, an alignment-free deep learning pipeline to identify viral reads and assemble viral contigs from tumor RNA-seq. Applied to 14 TCGA cancer types, the method uncovers expression of unexpected and divergent viruses not previously implicated in cancer and identifies human endogenous viruses (HERVs) whose expression associates with poor overall survival. This pipeline enables expanded characterization of the tumor virome and viral infections associated with clinical phenotypes.

Abdurrahman Elbasir, Ying Ye, Daniel E Schaffer, Xue Hao, Jayamanna Wickramasinghe, Konstantinos Tsilas, Paul M Lieberman, Qi Long, Quaid Morris, Rugang Zhang, Alejandro A Schaffer, Noam Auslander