Fig. 1

The landscape of viral expressions in human cancers. A Pipeline of viral sequence detection and downstream analyses. We downloaded 11,206 BAM files, expression profiles (fragments per kilobase million, FPKM) and clinical data including 33 cancer types from TCGA. For viral sequence detection, we aligned raw RNA-seq data in BAM files came from STAR with reference genomic sequences of human and 212 types of virus to detect viral sequences. We assembled transcripts with StringTie and the expression level of each transcript was normalized by TPM (transcripts per kilobase million). For transcripts of the same viral infection type, the maximum TPM value was selected as the final viral mRNA expression. Viral expression data, RNA-seq expression data and clinical data were then used for downstream statistical, genomic, transcriptional and immune analyses. B Fractions of viral infection in human cancers. The value of the fraction of viral infection is log10-transformed after adding a pseudo value of one. 20 types of cancer showed viral infections in over 1% of tumors, and 6 types of cancers showed viral infections in over 5% of tumors. C The landscape of viral expressions in human cancers. We showed viral expressions of all 744 infected samples from 25 types of cancer. Cancers are ordered according to the amount of infected samples. Cancers with highest amounts of viral infections are labelled, including CESC, LIHC, STAD, HNSC, LAML and COAD. The unlabeled cancers are ESCA, BLCA, OV, SKCM, READ, PAAD, UCEC, LGG, KIRC, SARC, PRAD, LUSC, KIRP, THCA, CHOL, DLBC, UCS, MESO, LUAD (from left to right). We detected 9 types of viral sequences in total, with 20 subtypes in HPV infections. Viral expressions were normalized into TPM values which are indicated with discrete colors notations. While most samples have a dominant type of viral infection, co-infections are also widespread. D Distributions of HPV, HBV, EBV and CMV infections in human cancers. Some cancers with low fractions are not labeled. The color of each sector indicates the average viral expression of tumors of that cancer type. E Distributions of viral expression levels. We plotted Kernel density plot for each virus (except for KSHV and MCV which only have a single sample). Viral expressions were log10-transformed and viruses are grouped based on their average expression levels