A novel approach to improve detection of somatic DNA variants in solid tumors by next-generation sequencing from FFPE samples.

Abstract

e22177 Background: Low frequency variant detection by sequencing is a highly desired goal for therapy selection in cancer especially the detection of actionable targets. The lower limit of detection using Sanger sequening is ~20% minor allele frequency (MAF). Deep sequencing of target genes using next generation sequencing (NGS) is gaining popularity. Formalin Fixed Paraffin Embedded (FFPE) tissue is the most common sample type in solid tumor histopathology. However, because the fixation process fragments DNA and damages it at varying frequencies, downstream processes can potentially misclassify modified bases and generate artifacts. We have developed a protocol that addresses both of these issues in a multiplex assay that involves deep sequencing using NGS of targets implicated in lung, gastric, colon, melanoma and ovarian cancers. Methods: The sample set includes 168 FFPE samples and 82 non FFPE samples. TruSeq Custom Amplicon technology was used to generate libraries for targets across 26 genes. Deep sequencing was done on a NGS platform (Illumina-MiSeq). Results: The DNA quality test, which surveys multiple genomic targets by qPCR, was an accurate determinant of DNA amplifiability and yielded a 99% sample success rate. A sensitivity of <5% MAF was achieved by sequencing at a depth of 1,000X for all targets. In order to differentiate true low frequency variants from fixation and other artifacts, our novel approach investigates each of the two DNA strands independently. The information is bioinformatically combined to distinguish true variants from artifacts. Testing of the FFPE samples with a 5% MAF cut off using the two strand approach reduced the potential false positive rate by ~ 40% when compared to information from only one strand of DNA. A comparative analysis of matched FF and FFPE sample showed that a high percentage of false positive calls were present even in the fresh frozen samples at this high level of sensitivity, if only using information from one strand. Conclusions: This protocol, efficiently and accurately detect low frequency variants by NGS in DNA extracted from FFPE tissues.