Fig. 8

Construction of eVIIS pipeline. The eVIIS pipeline includes several steps: (1) align RNA-seq data to human and viral reference genome sequences using STAR (version <= 2.5); (2) For viral expression detection: use StringTie (version <= 1.2.3) to assemble mapped reads and quantify viral expressions into TPM values. Samples with TPM over zero will be predicted as “Infected”, and samples with TPM equals zero will be predicted as “Nnon-Infected”; (3) For TIME status prediction: use featureCounts (version > = 1.5.0) to obtain read counts and normalize gene expressions into FPKM values. TIME prediction will be made based on the two models (LASSO regression and SVM). For a given sample, eVIIS first calculates its LF.Score using the LASSO regression model based on FPKM values of 30 LF-relevant genes. The sample will be classified as “Immune-Exclusion” if LF.Score < 81.03. Otherwise, it will be further evaluated by the SVM model as “Immune-Stimulation” or “Immune-Anergy”, using FPKM values of the 37 DEGs. The eVIIS pipeline is available at https://github.com/HuangLab-Fudan/eVIIS