329. IMPROVED TUMOUR PURITY AND PLOIDY ESTIMATION FROM LASER CAPTURE MICRODISSECTED ESOPHAGEAL ADENOCARCINOMA SAMPLES

Abstract

Abstract Large scale sequencing efforts by The Cancer Genome Atlas (TCGA), among others, have generated detailed genomic data for esophageal adenocarcinoma. However, samples with low cancer purity (i.e. high amounts of normal DNA) will limit accurate inference of copy number aberrations, a genomic hallmark of esophageal adenocarcinoma. This consequently restricts more advanced analysis such as quantifying intratumor heterogeneity or inferring tumor evolution. We have systematically collected biopsies from pre-treatment and post-induction esophageal adenocarcinoma tumors. Laser capture microdissection (LCM) was utilized to enrich for tumor cells for whole genome sequencing. Matching blood was utilized as germline control. Cancer purity and ploidy was estimated using a developed method tailored to the high genomic instability observed in EAC samples. This method uses both sequencing depth and minor allele frequencies to infer ploidy and purity. The esophageal adenocarcinoma dataset from TCGA were used as comparison. We achieved significantly higher tumor cellularity in surgical resection specimens (n = 8) compared to TCGA (71%+/−11 vs. 58%+/−18). Interestingly, tumor cellularity from biopsy specimens (n = 25) remained similar to TCGA despite LCM (65%+/−13), which is likely due to TCGA only using surgical resections that were pre-selected to have at least 60% cellularity, while we applied no histopathological cellular constraints on our samples. Estimated ploidy was similar between biopsy of primary tumors and surgical resection specimens compared to TCGA (2.67+/−0.58, 2.72+/−0.70, 2.64+/−0.75). There was a trend towards higher ploidy (3.07+/− 0.13) in metastatic biopsies (n = 3) but additional specimens are required. Laser capture microdissection of post-induction esophageal adenocarcinoma surgical resection specimens has resulted in significantly higher tumor purity in whole genome sequencing data. This will allow for improved inferences of phylogenetic trees between pre- and post- induction samples and quantification of subclonal copy number aberrations or intratumor heterogeneity.