Tumor Sequencing
A pathologist with specific expertise in thyroid tumors (N.M.Q.) examined the hematoxylin and eosin-stained tumor slides to identify representative areas of tumor suitable for molecular testing. DNA and RNA were extracted from FFPE tumor specimens according to the manufacturers’ protocols utilizing the QIAamp DNA FFPE Tissue Kit (Qiagen) and the RecoverAll Total Nucleic Acid Isolation Kit for FFPE (Ambion), respectively. DNA and RNA were quantified by Qubit (Thermo Fisher Scientific), with RNA undergoing an additional quality assessment by the quantitative PCR assay PreSeq (ArcherDx).
Of the original 40 cases selected for sequencing, 4 cases were excluded due to insufficient quantity and/or quality of nucleic acids. Therefore, 36 tumor samples with adequate quantity and quality of nucleic acids for sequencing were utilized for NGS library preparation. DNA NGS libraries were generated using a custom hybridization capture panel (NimbleGen/Roche) designed to capture all coding exons of 124 genes associated with pediatric solid tumors as well as the TERT gene promoter 20. RNA NGS libraries were generated using the Oncology Research Panel (ArcherDx) that utilizes Anchored Multiplex PCR chemistry to capture gene fusion events involving 75 recurrently rearranged cancer genes21. NGS libraries were sequenced 4- to 6-plex on an Illumina MiSeq using V2 and V3 chemistry, respectively, and de-mulitplexed using bcl2fastq v2.18.0.12 (Illumina), achieving approximately 2.2-4.4 million sequencing reads per sample. The resulting DNA FASTQ files were analyzed using a custom DNA bioinformatics pipeline that includes alignment by BWA v0.7.1222 (Wellcome Trust Sanger Institute) and deduplication and QC analysis by Picard v2.9.223. Variant calling was completed by NextGENe v2.4.1.2 (SoftGenetics) and Platypus v0.8.124 (Wellcome Trust Centre for Human Genetics) for the detection of SNVs and short indels, with variant annotation by VEP v.8925 (Ensembl). Copy number analysis was performed using CNVkit v0.9.326. CNVkit performs circular binary segmentation on the normalized GC and target capture density bias corrected log2 difference of binned read depths, using on-target (average bin-size = 300bp) and off-target reads (average bin-size = 500,000bp), between the tumor sample and reference pooled-normal peripheral blood samples to identify genome-wide regions of CNVs. Gene-level deletions and amplifications were then derived using these copy number segments. RNA FASTQ files were analyzed by the Archer Analysis suite v4.1.0.6 (ArcherDx) with additional custom wrapper and text-processing scripts for the detection of gene fusion events.