Abstract 571: Detection of somatic copy number alterations from on-target and off-target sequencing data

Abstract

Background: Several computational methods have been developed to identify copy number alterations (CNA) leading to or associated with cancer development and shown in recent studies to precede cancer diagnosis by many years. Current methods involving cell-free DNA (cfDNA) targeted sequencing data are based on the depth of coverage of on-target or off-target regions, whereas computational methods incorporating germline SNP information for making inferences on copy number alterations and tumor fraction remain underdeveloped. Methods: Using sequencing data from a large database of more than 100k clinical cell-free DNA (cfDNA) patient samples (Guardant Health, CA), we developed a probabilistic model to simultaneously normalize molecular coverage, segment the genome, predict copy number alterations, and estimate the tumor content in cfDNA samples, while accounting for mixtures of cell populations. The model is using off-target and on-target coverage. Copy number status, including loss of heterozygosity (LoH), is inferred in order to predict gene level somatic CNAs or genome wide instability/LoH. Results: We demonstrated the improvement from the off-target incorporation in three aspects. First, to estimate sensitivity improvement in detections of CNAs, we simulated deletions and amplifications of regions exceeding 40 Mb, using coverage and MAF variability observed in existing data. Combining coverage of on-target and off-target regions is expected to improve the LoD for detection of CNAs by 20%, when compared to CNA detection from on-target coverage. Next, we obtained samples from 15,618 cancer patients of different cancer types processed on GuardantOMNI® RUO and determined human leukocyte antigen (HLA) allele-specific copy number using this off-target assisted method. We observed a high prevalence (more than 15%) of LoH in HLA in bladder cancer, prostate cancer, NSCLC and HNSC, consistent with previous studies that HLA LOH is a common feature of several cancer types and diminishes immunotherapy efficacy. Finally, tumor fraction (TF) estimate was validated by comparing the TF against the maximum variant allele fraction of known oncogenic driver mutations in 6,000 cancer cases of various types. High concordance was observed in CRC samples (R2=0.75), gastric cancer (R2=0.63) and bladder cancer (R2=0.6), which suggest the use of this metric to better estimate tumor shedding levels in cfDNA in cases when driver mutations are not represented on a targeting panel. Conclusion: Our results show that probabilistic modeling of coverage data generated from both on-target and off-target cfDNA sequencing can detect gene specific or whole genome level somatic copy number alterations and LoH. This method may enable improvements in CNA detection accuracy, sensitivity, and specificity in plasma and provides more precise interrogation of LoH status and tumor fraction. Citation Format: Catalin Barbacioru, Han-Yu Chuang, Rebecca Nagy, Darya Chudova, AmirAli Talasaz. Detection of somatic copy number alterations from on-target and off-target sequencing data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 571.