A Comparison of Targeted Next Generation Sequencing from Paired Formalin-Fixed and Fresh Frozen Specimens
David H Spencer, Eric Duncavage, Robi D Mitra, Shashikant Kulkarni, Karen Seibert, Rakesh Nagarajan, Mark A Watson, John D Pfeifer. Washington University, St. Louis, MO
Background: Next-generation sequencing (NGS) has the potential to revolutionize molecular testing of cancer through comprehensive, unbiased, and inexpensive mutation detection. However, for these methods to be practical they must be robust to a diversity of specimen types, including formalin-fixed diagnostic tissue. We evaluated the performance and quality of targeted NGS in a series of paired formalin-fixed and fresh-frozen tissue to determine if formalin-fixed tissue is an adequate substrate for these platforms.
Design: Sequence data were generated for 15 pairs (30 specimens total) of case-matched fresh-frozen and formalin-fixed lung adenocarcinoma specimens. Indexed sequencing libraries were prepared with 1ug of DNA that passed ladder amplification and spectrophotometry quality-control assessment. Enrichment for 28 target genes was performed with custom capture reagents and enriched libraries were sequenced by multiplex sequencing. Variants were identified from mapped reads using a custom pipeline based on the Genome Analysis Toolkit. Read quality metrics and single-nucleotide variant calls were compared between sample types. For a subset of cases, calls from sample pairs were compared to previously-generated array-based genotype data.
Results: Multiplex sequencing produced 2,000-10,000-fold coverage of the target region across all samples. There were small, but statistically-significant differences in insert size (179 bp (fixed) vs 225 bp (frozen); P<10-7), aligned bases per read (96 bp (fixed) vs. 98 bp (frozen); P<10-5), and the fraction of unique reads (22% (fixed) vs 31% (frozen); P<10-3). However, the mean number of high-quality bases per read, mean error rate, and proportion of positions with high-quality basecalls were not significantly different. Basecall concordance between paired samples was >99.99%, with an average of only 5 discordant positions over the ∼300 kilobase-pair target interval per case. The concordance between basecalls and array-based genotypes was >98% for each specimen type.
Conclusions: We have shown that next-generation sequencing using DNA from formalin-fixed specimens yields comprehensive and accurate sequence information that is comparable to that from fresh-frozen tissue. These results indicate that NGS technology can be readily integrated into the standard surgical pathology workflow and demonstrate the feasibility of large-scale sequencing of formalin-fixed surgical pathology specimens for basic science, translational research, and clinical trials.
Category: Special Category - Pan-genomic/Pan-proteomic approaches to Cancer
Tuesday, March 20, 2012 1:00 PM
Platform Session: Section H1, Tuesday Afternoon