ABSTRACT
Next-generation sequencing provides an opportunity to detect viral species from RNA-seq data of human tissues, but existing computational approaches do not perform optimally on clinical samples. We developed a bioinformatic method called VirTect for detecting viruses in neoplastic human tissues using RNA-seq data. Here, we used VirTect to analyze RNA-seq data from 363 head and neck squamous cell carcinoma (HNSCC) patients and identified 22 human papillomavirus (HPV)-induced HNSCCs. These predictions were validated by manual review of pathology reports on histopathologic specimens. VirTect showed better performance in recall and accuracy compared to the two existing prediction methods, VirusFinder and VirusSeq, in identifying viral sequences from RNA-seq data. The majority of HPV carcinogenesis studies thus far have been performed on cervical cancer and generalized to HNSCC. Our results suggest that carcinogenesis of HPV-induced HNSCC and other cases of HNSCC involve different genes, so understanding the underlying molecular mechanisms will have a significant impact on therapeutic approaches and outcomes. In summary, RNA-seq together with VirTect can be an effective solution for the detection of viruses from tumor samples and can facilitate the clinicopathologic characterization of various types of cancers with broad applications for oncology.