BIOM-12. ULTRA-DEEP WHOLE GENOME SEQUENCING UNCOVERS A HIGH PREVALENCE OF GLIOBLASTOMA-DERIVED CELL-FREE DNA IN PLASMA

Abstract The detection of glioblastoma-derived circulating cell-free DNA (ccfDNA) has proven formidable because tumor-specific mutations are hidden among the abundance of ccfDNA from healthy cells. While capture-enrichment panels provide the necessary read depth to detect low-frequency variants, searches are constrained to a narrow range of targets and largely limited to single nucleotide variants (SNVs). Here, we use ultra-deep whole genome sequencing (WGS) to broaden the search for somatic mutations across the entire genome. In addition, a kmer-based reference-free variant caller, which mitigates the bias and errors that accrue during alignment, is used to extend the mutation search beyond SNVs to include insertions/deletions (INDELS) and structural variants (SVs) such as large INDELS ( >20 bp) and tandem duplications. In tumor DNA and ccfDNA, we achieve a WGS (GRCh38) mean read depth of 263X and 225X, respectively. WGS of matched buffy coat DNA and sequencing reads from the 1000 Genome Database were used to eliminate germline polymorphisms, sequencing artifacts, and common variants. We detected 20,488 somatic mutations in tumor DNA of which 17.3% affected a coding sequence, 31.9% were INDELS, 91 were large INDELs, and 18 were tandem duplications. A tumor-informed search identified 489 of these somatic mutations in ccfDNA which included a large INDEL and a similar portion of mutations affecting a coding sequence. Using an unguided search for glioblastoma-derived ccfDNA uncovered an additional 8,794 somatic mutations of which 16.6% involved a coding sequence, 60.7% were INDELS, 30 were large INDELS, and 3 were tandem duplications. The high prevalence of somatic mutations discovered in ccfDNA using an unbiased approach highlights the low yield of tumor-informed searches in spatially genetically heterogeneous cancers like glioblastoma. Our findings demonstrate that ultra-deep WGS detects glioblastoma-derived ccfDNA by broadening the search for variants across the entire genome and through the inclusion of INDELS and SVs.