Streamlined NGS Analysis for CNV Identification

Application Note Authors Alejandra Bernardini, Agilent Technologies Vincenzo Lupo, Agilent Technologies Ivan Lesende, Agilent Technologies Joachim de Schrijver, Agilent Technologies Christine Hörnlein, Agilent Technologies Introduction Many genetic laboratories utilize custom panels for genetic diagnosis of germline variants of different diseases, such as the diagnosis of hereditary cancer (HC). While identifying single nucleotide variations (SNVs) and insertions and deletions (indels) from next-generation sequencing (NGS) data has become routine and well established, it is still challenging to accurately call germline exon-level copy number variations (CNVs) from small panelcaptured libraries. Meanwhile, comparative genome hybridization (CGH) and multiplex ligation-dependent probe amplification (MLPA) are more conventional and established methodologies for CNV detection. Agilent provides the flexibility to customize sequencing panels that enables the detection of genomic alterations and delivers the technology tailored to meet user-specific needs for NGS target enrichment. The Agilent NGS end-toend workflow solution includes web-based software to create a sequencing design (Agilent SureDesign), library preparation (Agilent SureSelect Library Prep and Enrichment), automation platforms (Agilent Magnis NGS Prep system and Agilent Bravo NGS workstation), sample and library quality control (QC) (Agilent TapeStation systems), and, ultimately, solutions for data analysis and interpretation of genomic variants (Agilent Alissa Reporter). The Alissa Reporter platform is an intuitive and streamlined, cloud-based NGS secondary analysis software-as-a-service (SaaS) solution that delivers high-performance variant detection (that is, SNVs, indels, and CNVs) with integrated genome browsing, a built-in QC dashboard, and seamless connectivity to SureDesign software. Integrating the NVIDIA Clara Parabricks platform into the secondary analysis pipelines enables users to benefit from fast, panel-specific, and cost-efficient data analysis solutions. Alissa Reporter allows complete automation of data upload, analysis, and export, through your Amazon Web Services account. CNV Calling Performance of Custom Panels in Alissa Reporter 2 Materials and Methods Sixteen genomic DNA (gDNA) samples with known CNVs, obtained from cancer patients’ whole blood, were used to test CNV performance in Alissa Reporter. Pathogenic CNVs were previously validated using either MLPA or CGH methods. gDNA samples were subjected to targeted enrichment by using different Agilent technologies and workflows for each case, including enzymatic or transposase-based shearing methods for DNA fragmentation. Library preparation was manually performed using the Agilent SureSelect QXT system or automatically performed using the Magnis NGS Prep system and Magnis SureSelect XT HS reagent kit, or the Bravo NGS platform and Agilent SureSelect XT Low Input reagent kit (Table 1). The capture probe design includes all coding regions, non-coding variants, and promoters of interest associated with hereditary cancer genes (curated by clinical experts of the different hospitals), sex determination, and sample tracking single nucleotide polymorphisms (SNPs). Target-enriched libraries were sequenced on an Illumina sequencing platform. Demultiplexed FASTQ files from sample libraries were uploaded to Alissa Reporter software. Table 1. Samples were prepared with different SureSelect library preparation workflows. This application note describes the high performance of Alissa Reporter in detecting a broad range of CNVs from different hospital patient samples that were previously validated by orthogonal methods. The targeted regions associated with hereditary cancer genes were enriched for sequencing using Agilent SureSelect XT HS and QXT reagent kits, captured-based technologies. The Alissa Reporter CNV calling algorithm enables exonlevel copy number determination and uses relative coverage differences of one sample against other similar samples to calculate log2 ratios and eventually make CNV calls. In more detail, CNV calling is performed using the read data in the BAM file following positional deduplication and removal of reads with low mapping quality. CNVs are called as deviations in the read depth of the target sample interval compared to the expected (diploid) read depth, as estimated from a set of unrelated samples within the same sequencing run batch. Product Name SureSelect XT HS SureSelect XT-LI SureSelect QXT DNA Input Shearing Automation Samples per run 10-200 ng Enzymatic Magnis system 8 samples 10-200 ng Enzymatic Bravo NGS 96 samples 50 ng Transposase-based No Results Data quality Alissa Reporter’s QC dashboard was used to assess sample quality and performance. As a representative example, Figure 1A displays QC tiles of a run where samples S1 to S3 (Table 2) were analyzed, and the pop up from total raw reads shows median, average, minimum, and maximum values of raw reads. Figure 1B displays a pop up from the coverage QC tile, showing plot coverage of all samples in a run that includes samples S11 to S16. A heatmap sample correlation plot (Figure 1C) shows no outliers and a good correlation between samples in a representative run. For the overall set of samples, the average uniformity (fraction of targeted bases at 20X coverage) was 0.95 (± 0.049 and average coverage was 237.1 ± 50.98). Variant calling Alissa Reporter detected all the previously validated CNVs in the 16 tested samples (that is, one or two CNVs per sample) (Table 2). Fifteen out of 16 CNVs were deletions, and one CNV was an amplification in exon 8 of MSH2 in sample S16 (Table 2). Among the deletions, one spans a large region including entire genes (RB1 and BRCA2 in sample S14). One sample contained two CNV deletions in a single gene, the first spanning exon 1 to 11 of APC and the second affecting exon 15 of same gene (sample S15, Figure 2). The remaining 13 CNV deletions ranged from several exons to one exon, either affecting only one gene (MSH2, PALB2, BRCA2, STK11, MUTYH, and VHL) or two consecutive genes (EPCAM-MSH2 and EPM2AIP1-MLH1) (Table 2). As a representative example, an EPCAM-MSH2 deletion spans 16 exons (sample S7, Figure 3), while the deletion in STK11 targeted only one exon (sample S9, Figure 4). Leveraging the functionality of synchronizing views, it is possible to see a loss of heterozygosity (LOH) in the deleted regions as detected for sample S14 where the entire BRCA2 is deleted in heterozygosis (Figure 5). BRCA2 variants overlapping the deletion had all 100% AF and average quality of 2621 (range 662-5683). 3 A B C Figure 1. QC metrics; figures derived from Alissa Reporter v1.1. A. QC tiles for an example run showing read QC metrics: the number of reads tile is popped up showing detailed metrics. B. QC tiles for an example run showing enrichment metrics: the average depth in target region bar graph is popped up. C. The CNV heatmap represents Pearson correlation coefficient between sample pairs in an example run. Sample_name CNV type Gene Position CNV size (bp) N exons Chemistry Automation Confirmation S1 Deletion MSH2 chr2: 47,702,081 8092 5 XT-HS Magnis system MLPA S2 Deletion PALB2 chr16: 23,632,559 20054 10 XT-HS Magnis system MLPA S3 Deletion BRCA2 chr13: 32,918,527 2642 2 XT-HS Magnis system MLPA S4 Deletion BRCA2 chr13: 32,918,527 2642 2 XT-LI Bravo NGS MLPA S5 Deletion MSH2 chr2: 47,697,971 12202 6 XT-LI Bravo NGS MLPA S6 Deletion EPCAM,MSH2 chr2: 47,612,114 18635 2 - 1 XT-LI Bravo NGS MLPA S7 Deletion EPCAM,MSH2 chr2: 47,600,450 72498 8 - 8 XT-HS Magnis system MLPA S8 Deletion EPM2AIP1,MLH1 chr3: 37,029,990 8359 1 - 3 XT-HS Magnis system MLPA S9 Deletion STK11 chr19: 1,206,804 508 1 XT-HS Magnis system MLPA S10 Deletion MSH2 chr2: 47,641,301 2371 2 XT-HS Magnis system MLPA S11 Deletion VHL chr3: 10,188,115 3565 2 QXT Manual MLPA S12 Deletion MUTYH chr1: 45,794,900 4412 14 QXT Manual MLPA S13 Deletion VHL chr3: 10,183,498 4905 2 QXT Manual MLPA S14 Deletion BRCA2,RB1 chr13: 32,890,523 16163868 whole genes QXT Manual CGH S15 Deletion APC chr5: 112,043,211 114612 11 QXT Manual MLPA S15 Deletion APC chr5: 112,170,561 388 1 QXT Manual MLPA S16 Amplification MSH2 chr2: 47,672,560 364 1 QXT Manual MLPA Table 2. Exon-level CNV calling of all tested CNVs. 4 Discussion NGS is increasingly being adopted for genomic testing. Although SNV and indel detection performs well on NGS data, CNV detection at the exon-level still poses challenges.1 Alissa Reporter offers a friendly and easy-to-use NGS data analysis platform that provides rapid sample results directly from FASTQ files. Although the compared samples came from different labs and were subjected to diverse handling approaches such as various library preparation chemistries and sequencing instruments, the QC investigation, leveraging Alissa Reporter’s QC dashboard, revealed only minor data quality biases. In our comparison study, Alissa Reporter detected a broad range of CNVs previously validated by orthogonal methods from patient samples. It was also able to call all previously validated CNVs, including single exon CNVs and complex CNVs such as two CNVs in one gene. The heatmap feature in Alissa Reporter’s QC dashboard displayed no sample outliers which could interfere with building the reference set and supported the conclusion there was no distinguishable effect on performance between the diverse Agilent chemistries (QXT and XTHS) or workflows (manual or automated with either Bravo or Magnis). Conclusion Alissa Reporter's CNV calling pipeline demonstrates robust and accurate calling performance over a wide range of hereditary cancer datasets down to exon-level resolution. Alissa Reporter precisely detected all CNVs previously found through other methods. This highly sensitive variant calling performance emphasizes Alissa Reporter’s robustness and accuracy and shows its readiness for adoption into a laboratories routine workflow. Figure 2. Combined deletion in APC spanning exon 1 to 11 and exon 15 (sample S15 in Table 2). Red track indicates CNV deletion; figure derived from Alissa Reporter v1.1. Figure 3. A deletion spanning EPCAM 3’ region (8 exons) and MSH2 5’ region (8 exons) resulted in a gene fusion (sample S7 in Table 2); figure derived from Alissa Reporter v1.1. Figure 4. A deletion in STK11 targeted the first exon (sample S9 in Table 2); figure derived from Alissa Reporter v1.1. Figure 5. Entire deletion of BRCA2 (sample S14 in Table 2). Synchronizing view functionality of Alissa Reporter between CNV viewer zooming in 10 out of 27 exons of BRCA2 deleted in heterozygosis (above) and SNV pileup showing SNPs with AF of 1 indicating LOH (below); figure derived from Alissa Reporter v1.1. References 1. Zarrei, M.; MacDonald, J. R.; Merico, D.; Scherer, S. W. A Copy Number Variation Map of the Human Genome. Nat. Rev. Genet. 2015, 16 (3), 172–183. https://doi.org/10.1038/nrg3871. Acknowledgements Technical support and FASTQ files were kindly provided by: – Genetic Unit. University Hospital Nuestra Señora de Candelaria. Tenerife (Spain) – Genetic Unit. Hospital Universitari I Politècnic La Fe. Valencia (Spain) – Biochemistry and Clinical Genetics center. Clinic University Hospital Virgen de la Arrixaca. Murcia (Spain) – Eurofins-Megalab. Molecular Diagnostic Department. Madrid (Spain) Technical support was also kindly provided by Molecular Genetic. University General Hospital of Elche. (Spain) www.agilent.com Research Use Only. Not for use in diagnostic procedures. PR7001-2393 This information is subject to change without notice. © Agilent Technologies, Inc. 2021-2024 Published in the USA, May 1, 2024 5994-6585EN