Classification of intron retention associated variants using tumor and matched-control exome sequencing data

For each transcriptome sequencing data having detected IRAVs, we first checked whether the corresponding tumor and matched control exome sequencing data is available in the Genomic Data Commons. Here, we focused on the sample type of “Primary Tumor,” and “Primary Blood Derived Cancer - Peripheral Blood” (as well as “Metastatic" in cases when the cancer type is TCGA-SKCM) for tumor and “Solid Tissue Normal” and “Blood Derived Normal” for matched control. When both tumor and matched control were obtained, we downloaded part of the BAM file around the detected IRAVs using the BAM Slicing API. Then, we measured sequencing depths, the numbers of IRAV supporting reads, and the ratios of IRAV supporting reads for the IRAVs and classified them into “germline,” “somatic,” “somatic or germline,” “ambiguous,” and “false positive” according to the flowchart presented in Fig. 2a.