INDELs in Cancer Detected by Targeted Clinical Next Generation Sequencing
Jennifer Sehn, David Spencer, Haley Abel, Eric Duncavage. Washington University in St Louis, St. Louis, MO
Background: The spectrum of insertions and deletions (INDELs) ranging from 1bp to ∼1kb is poorly characterized in cancer. Such INDELs can be detected using next generation sequencing (NGS) techniques, with higher resolution than karyotyping, and without prior knowledge of the area of mutation as required for PCR. However, INDEL detection by NGS for targeted oncologic gene panels in routine clinical care remains uncharacterized.
Design: Targeted NGS for 27 cancer genes from 97 consecutive tumor specimens (52 lung, 14 pancreas, 9 colon/rectum, 4 breast, and 18 other) was performed in multiplex using 2x101bp paired-end reads at >1000x coverage. Sequence data were aligned to hg19 and coding region INDELs were identified using Pindel. The accuracy of the INDEL calls was determined by visual inspection of the raw sequence alignments.
Results: Visual review of sequence reads in areas containing INDEL calls revealed that multiple distinct variant annotations were often made by the variant calling software when a single INDEL occurred within a region of low sequence complexity. For these INDELs, the flanking DNA sequences were the same, so the start site of the INDEL was ambiguous, leading to multiple annotations for the same sequence alteration; the end result was an artificially decreased number of supporting reads for each INDEL call. When INDEL calls that by visual inspection represented the same sequence alteration were combined into a single annotation, a total of 292 coding-region INDELs with at least 10 supporting reads were identified (average 5/case, range 0-76/case), 147 of which were insertions and 145 deletions. The average INDEL length was 16 bp (range 1bp-8029bp). Breast carcinomas had the highest average number of INDELs/case (7), followed by colorectal (4), lung (3) and pancreas (2). The most commonly mutation gene was MLL (52).
Conclusions: Targeted NGS of cancer genes shows a constellation of INDELs that are not readily identifiable by other laboratory techniques. Careful review of INDEL calls made by open-source variant calling packages reveals an inherent tendency to annotate a single INDEL with an ambiguous start site as occurring at multiple start sites, with a resulting decrease in the number of supporting reads for each call. Consequently, INDELs with low numbers of supporting reads are filtered out of NGS analyses, contributing to underrepresentation of this potentially clinically relevant class of mutations.
Category: Special Category - Pan-genomic/Pan-proteomic approaches to Cancer
Monday, March 4, 2013 9:30 AM
Poster Session I Stowell-Orbison/Surgical Pathology/Autopsy Awards Poster Session # 277, Monday Morning