Abstract 5459: Combining targeted sequencing and ultra-low pass WGS for accurate SCNA detection on tumor sample of biased aneuploidy

Abstract

Abstract Background: Somatic copy number alteration (SCNA) as a common genomic abnormality in tumor tissue has been routinely screened in cancer research due to its role in tumorigenesis. One popular approach of genome scale SCNA detection is ultra-low pass WGS. However, when dealing with tumor sample of highly biased aneuploidy, this method makes biased normalization of sequencing throughput and consequently yields inaccurate SCNA result. Recent pan-cancer study revealed that whole genome doubling (WGD), a typical example of severe aneuploidy, occurs in 56% of tumor samples, much higher than previous estimation. This observation highlights the imminent need for developing more accurate SCNA detection method especially when dealing with tumor with high aneuploidy. Here, we present a targeted sequencing guided SCNA detection method that allows accurate SCNA detection through improved normalization method. Method: Both tumor and normal tissue samples were subjected to targeted sequencing (genomic coverage: 300kb) and 2X ultra-low pass WGS. Genomic regions were binned and corresponding depth was calculated. GC bias was corrected using LOWSS regression. Single nucleotide polymorphism (SNP) was detected from targeted sequencing. We then selected all genomics regions that have stable sequencing depth as well as stable VAF of heterozygous SNP. Among them, regions of lowest depth were defined as diploid and used for sequencing throughput normalization. To compare SCNA results from different methods, we defined CN-VAF conflict index to represent the deviation of VAF of heterozygous SNP from diploid region. Results and conclusion: We applied our method to a total number of 29 NSCLC samples. In comparison with CNVkit and ControlFreeC results, 18 tumor samples shown consistent SCAN results. For the rest of 11 samples, CNVkit and ControlFreeC resulted in overall higher SCNA across genome comparing to our method. Correlation analysis shown that the inconsistency of SCNA status between methods positively correlates with the difference of baseline estimation (Pearson coef: CNVkit: 0.93, ControlFreeC: 0.91). For samples with inconsistent SCNA results, CN-VAF conflict index indicated that our method made more accurate SCNA estimation through improved baseline estimation. In addition, experimental validation using ddPCR shown consistent SCNA results with our method. It is worth mentioning that the key step of our method is the estimation of baseline depth which depends on the length of genomic region, the amount of SNP within the genomic region and the fluctuation of both depth and SNP VAF in the region. If sufficient tumor samples with experimentally validated aneuploidy are available, machine learning approach may make further improved estimation. Lastly, it is worth mentioning that this article is not intended to propose using both targeted sequencing and ultra-low WGS in routine NGS experiment, but rather as a complementary approach to evaluate the heterogeneity of aneuploidy tumor. Citation Format: Cheng Yan, Weihua Guo, Yufei Yang. Combining targeted sequencing and ultra-low pass WGS for accurate SCNA detection on tumor sample of biased aneuploidy [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5459.