Comprehensive clinical tumor RNA sequencing for children with cancer.

e15067 Background: Cancer is the number one cause of death by disease for children in the United States. The underlying cause for most childhood cancers is unknown and unlike most adult cancers, childhood cancers harbor fewer coding mutations compared to adult cancers and are often driven by epigenetic deregulation. The functional impact of both genetic and epigenetic abnormalities can be detected from the analysis of tumor RNA. While tumor RNA profiling has entered several clinical trials, it is not utilized comprehensively and there is limited availability of clinical-grade RNA-Seq assays. We have published preliminary evidence of the utility of using research tumor RNA information in the pediatric oncology clinic. The next step is to develop an RNA-Seq assay that identifies gene fusions, expressed mutations, and important gene expression abnormalities in pediatric tumors. Methods: To validate our RNA-Seq assay pipeline, we have started with fusion detection and compared our results with a CLIA-validated DNA-based assay. We used RNA-Seq data generated from three fusion-positive cell lines to define the minimum of Mapped, Exonic, Non-Duplicate (MEND) reads necessary to accurately identify known fusions. Results: We found that the fusion, which has moderate expression (above 3 log2(TPM+1)), was consistently detectable in subsets containing at least 1 million MEND reads (Table). We next evaluated the precision of this finding. The fusion was detected in each of 21 additional subsets of 2 million MEND reads with replacement, but in only 71% of 21 subsets of 1 million MEND reads. In RNA-Seq data generated from tumor samples preserved with FFPE, we compared our RNA-based fusion findings with structural variants detected with the DNA-based gene panel test UCSF 500. Seven positive results were concordant, and one negative result was concordant. Two negative findings were positive by UCSF 500. Evaluation of these discordant results is ongoing, but the expression data indicates that the RNA result for one, a TMPRSS-ERG fusion, is accurate. To incorporate gene expression into our assay, we assessed the precision of gene expression measurements. In a preliminary dataset of three technical replicates, we found that the 95% confidence interval of expression values greater than 3 log2(TPM+1) is 4.5% of the measured value. Comprehensive clinical validation of our protocol will involve additional previously-characterized cancer samples. Conclusions: In summary, these data establish and validate key analytical parameters necessary for successful implementation of RNA-Seq analysis in the pediatric oncology clinic and advance our ultimate goal of bringing the most innovative molecular diagnostic methods to every child with cancer. [Table: see text]