- Email: [email protected]
Deletions, and Fusion Genes
Journal Pre-proof Clinical Validation of a Myeloid Next-Generations Sequencing Panel for Single Nucleotide Variants, Indels, and Fusion Genes Iyare Izevbaye, Li Ying Liang, Cheryl Mather, Soufiane El-Hallani, Remegio Maglantay JR, Lalit Saini PII:
S1525-1578(19)30409-X
DOI:
https://doi.org/10.1016/j.jmoldx.2019.10.002
Reference:
JMDI 845
To appear in:
The Journal of Molecular Diagnostics
Received Date: 5 June 2019 Revised Date:
11 September 2019
Accepted Date: 7 October 2019
Please cite this article as: Izevbaye, I, Liang, LY, Mather, C, El-Hallani, S, Maglantay JR, R, Saini L, Clinical Validation of a Myeloid Next-Generations Sequencing Panel for Single Nucleotide Variants, Indels, and Fusion Genes, The Journal of Molecular Diagnostics (2019), doi: https://doi.org/10.1016/ j.jmoldx.2019.10.002. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Copyright © 2019 Published by Elsevier Inc. on behalf of the American Society for Investigative Pathology and the Association for Molecular Pathology.
Clinical Validation of a Myeloid Next-Generations Sequencing Panel for Single Nucleotide Variants, Indels, and Fusion Genes Iyare Izevbaye,* Li Ying Liang,* Cheryl Mather,* Soufiane El-Hallani,* Remegio Maglantay JR,* and Lalit Saini† From the Department of Laboratory Medicine and Pathology,* University of Alberta Hospital, Edmonton, Alberta; and the Department of Medicine,† University of Alberta and Cross Cancer Institute, Edmonton, Alberta, Canada
Corresponding author: Iyare Izevbaye, Department of Laboratory Medicine and Pathology, 8440 112 Street NW, University of Alberta Hospital, Edmonton, AB, Canada T6G2B7, Tel: 780407-8025 Fax: 780-407-8599; Email: [email protected]
Short running title: Validation of a myeloid NGS panel
Funding: Supported by Marshall Eliuk Fund and University Hospital Foundation (SFR1404), University of Alberta, Alberta Public Laboratories, Edmonton; grant recipient: II.
Disclosures: I.I. has received honorarium for consultation/advisory role for Novartis, Pfizer, Roche, AstraZeneca, Bayer, Precision Rx-Dx.
Abstract Myeloid neoplasms are a heterogenous group of neoplasms including acute myeloid leukemia (AML), myeloproliferative neoplasms, myelodysplastic syndrome, and myeloproliferative neoplasms / myelodysplastic syndrome. Genetic abnormalities are used as diagnostic, prognostic, and predictive biomarkers in patients with these diseases. Here, we describe the clinical validation of the Oncomine Myeloid Research (OMR) next-generation sequencing panel that interrogates for 40 genes and 29 fusion genes commonly seen in myeloid neoplasms. Our validation set of 77 DNA samples included acute and chronic myeloid neoplasms with 91 single nucleotide variants and small indels. The 71 RNA samples from patients with AML included most of the AML-defining translocations. The OMR on the Ion Torrent S5 platform shows good performance in terms of depth of coverage, on-target reads, and uniformity. The panel achieved 91.3% and 100% concordance with reference DNA and RNA samples, with a clinical sensitivity and specificity of 96.7% and 100% for DNA, and 99.8% and 100% for RNA, respectively. Precision and reproducibility were 100% and the lower limit of detection was generally 5% VAF for DNA and 2-log reduction from initial value for RNA fusion genes. In conclusion, the OMR panel is a highly accurate and reproducible next-generation sequencing panel for the detection of common genetic alteration in myeloid neoplasms.
Introduction Next-generation sequencing (NGS) is an increasingly important diagnostic tool in the characterization of myeloid malignancies. For many years, World Health Organization (WHO) classification of acute myeloid leukemia (AML) has established the importance of recurrent cytogenetic and molecular features for diagnosis and prognosis [1]. According to the WHO scheme, class-defining cytogenetics and molecular features include recurrent gene rearrangements for t(15;17), t(8;21); inv(16)-t(16;16), t(9;22), t(9;11), t(6;9), inv(3)-t(3;3), t(1;22); and mutated NPM1, mutated CEBPA, and the provisional entity of mutated RUNX. FLT3 is also described as a poor prognostic marker in AML. With widespread application of genomic technology to the study of AML, many new molecular biomarkers have been shown to be prognostic and therapeutic targets. In a study of 200 adult AML cases, a total of 23 genes were significantly mutated. These genes could be placed in distinct functional categories including transcription factor fusions, tumor suppressor genes, DNA methylation-related genes, signaling genes, chromatin-modifying genes, myeloid transcription factor genes, cohesin-complex genes and spliceosome-complex genes. Functional relationships and genetic interplay between these genes were demonstrated by patterns of cooperation and mutual exclusivity seen in the mutations [2]. In another study of 1,540 patients, the authors argue for a genomic classification of AML based on driver mutation causality, an approach with clear implications for targeted therapy. Driver mutations were identified in 76 genes or genomic regions, with two or more drivers identified in 86% of patients. The authors suggested three genomic categories for AML including chromatin-spliceosome, TP53-aneuploidy, and IDH2R172 mutations. The genomic
categories and the presence or absence of other driver mutations were significant determinants of clinical outcomes [3]. Chronic myeloid neoplasms are a heterogenous group of myeloid neoplasms comprising myeloproliferative neoplasms (MPN), including chronic eosinophilic leukemia (CEL), myelodysplastic syndrome (MDS), and MPN/MDS. These myeloid entities are also diagnosed and classified based on molecular features. Chronic myelogenous leukemia (CML) is the classical molecularly defined neoplasm, in which the diagnostic t(9;22) translocation harbors comprehensive properties for diagnostic, predictive, monitoring, and resistance detection in the diagnosis and evolution of the disease. Non-BCR/ABL1 MPN also have defining genetic profiles with mutations in JAK2, CALR, and MPL [4-6]. Recurrent gene rearrangements defining the myeloid and lymphoid neoplasms with eosinophilia include abnormalities of PDGFRA, PDGFRB, FGFR and the newly identified PCM1-JAK2 translocation [1]. With the identification of new biomarkers and genetic targets, genomic characterization is becoming routine in the diagnosis and management of chronic myeloid neoplasms. NGS profiling has demonstrated utility for diagnosis, prognostic stratification, and therapeutic choice in myeloid neoplasms [7-10]. For example, SF3B1 mutations show correlation with MDS with ring sideroblasts9. Mutations in a number of genes such as TP53, EZH2, ETV6, RUNX1, and ASXL1 are independent prognostic predictors of reduced overall survival in MDS. NGS testing is frequently contributory in diagnostically difficult myeloid neoplasms by establishing clonality in borderline cases and differentiating neoplastic from reactive conditions [10]. This is particularly useful when morphologic and other routine molecular and cytogenetic tests are inconclusive.
Many different NGS myeloid panels including commercial and custom laboratorydeveloped tests (LDT) have been developed [11-14]. Consensus is still developing about the role of these biomarkers and which must be included in a myeloid panel. Recently, AMP released practice guidelines recommending a basic number of medically relevant genes that should be included in a myeloid panel [15]. Due to the multiplicity of genetic and cytogenetic alterations in myeloid malignancy and the need for prompt clinically required TAT, care must be taken in choosing an appropriate NGS assay that must meet these clinical requirements. Other important considerations include the local laboratory setting, the availability of technical manpower and process automation; and a sufficiently comprehensive spectrum of genes in the panel to cater for the variety of myeloid neoplasms to be managed in the clinical setting the laboratory serves. In this study, we report the clinical validation of the Oncomine Myeloid Research (OMR) panel on the Ion Torrent S5 (Thermo Fisher Scientific, San Francisco, CA). We studied the feasibility of a myeloid panel that is capable of providing comprehensive genomic information including recurrent cytogenetic analysis, SNVs, and small insertions-deletions (indels) for the diagnosis of a multiplicity of acute and chronic myeloid neoplasms. Examples of such entities having these diverse genetic aberrations include AML, MPN, MDS, and MDS/MPN. We assessed the performance of the OMR on the Ion Torrent S5 using samples previously characterized on the Illumina Trusight Myeloid panel (TMP; Illumina, San Diego, CA) and an Illumina custom panel designed by a referral laboratory, British Columbia Cancer Agency (BCCA).
Materials and Methods This study was reviewed and approved by the Health Research Ethics Board of Universities of Alberta and Calgary. OMR Panel content and genes/regions The OMR panel consists of RNA- and DNA-based gene panels with 526 DNA and 700 RNA amplicons. The average DNA amplicon size is 279 base pairs (bp) with an insert size of 227bp. The average RNA amplicon size is 143bp with an insert size of 99bp. The panel targets the complete exonic regions of 17 genes, exonic hot spots of 23 genes, 29 fusion genes, five expression genes, and five control genes (Table 1). Validation Samples DNA and RNA were extracted on the Qiasymphony instrument (Qiagen, Hilden, Germany) using the QiaSymphony kit from previously well-characterized peripheral blood (26) or bone marrow (122) samples. Nuclei acid was quantified by Nanodrop (Thermo Fisher Scientific) and Qubit HS Assay Kit (Thermo Fisher Scientific). Samples were run unpaired for DNA and RNA analysis. A total of 77 DNA samples were analyzed, comprising of 67 samples previously referred out for NGS analysis on the TMP panel or a custom illumina NGS myeloid panel; and an additional 10 DNA samples previously characterized for specific mutations (CALR, FLT3, KIT, and CEBPA) by in house methods. These additional samples were chosen to assess the capacity to sequence challenging genomic regions with high GC contents, egCEBPA and the limiting length of indels, egCALR and FLT3-ITD. The samples consisted of 49 samples with well characterized mutations and 28 negative samples with no pathogenic mutations. Benign, likely benign, variants of
undetermined significance, and low confidence variants were designated as no pathogenic mutation detected. The samples had a range of gene variants from 0 to 7. The variants included SNVs and indels of different sizes from ≤5, 5 to 40, >40bp (Table 2). The sample cohorts harbored variants in a spread of 26 unique genes characteristic of myeloid neoplasm. The diagnostic classes for DNA analysis included AML (42%), MPN (22%), MDS (6%), MDS/MPN (6%), clonal cytopenia of undetermined significance (CCUS; 4%), and others (20%).
A total of 71 RNA samples were analyzed, comprising of 45 samples with an abnormal fusion gene and 26 with normal or fusion negative karyotype. Two samples failed NGS analysis and were excluded from further assessment, giving a total of 43. The samples had been previously characterized by karyotyping and/or FISH. This selection covers seven of eight of the defined AML with recurrent genetic abnormalities and a FIP1L1-PDGFRA gene rearrangement (Table 3). The diagnostic classes for RNA analysis included AML (44.2%), acute promyelocytic leukemia (APL; 9.3%), acute lymphocytic leukemia (ALL; 4.6%), CML (39.5%), and CEL with PDGFRA gene rearrangement (2.3%). This selection covers all common AML with recurrent genetic abnormalities.
Commercial controls included in the analysis comprise Seraseq Myeloid Mutation DNA Mix (Seracare, Milford, MA; product number 0710-0408), having 23 clinically–relevant DNA mutations across 16 genes and the SeraSeq Myeloid Fusion RNA Mix (Seracare, product number 0710-0407) having nine RNA fusions. Well characterized Coriell wild-type DNA control (NA12878) (Coriell Institute of Medical Research, Camden, NJ) and IVS (Invivoscribe, San
Diego, CA) negative RNA control (IVS 0035) were used for preparation of dilutions for sensitivity testing for DNA and RNA panels, respectively, and for accuracy studies.
Library preparation, templating, and sequencing OMR panel library preparation was performed by manual and automated processes according to the manufacturer’s instructions (Thermo Fisher Scientific). Briefly, barcoded libraries were generated from 10 ng of sample gDNA and RNA. Targets were amplified using an RNA primer pool and two DNA primer pools using highly multiplexed PCR amplification. The amplicons were partially digested with Fupa enzyme followed by ligation of unique barcode adapters for each library. The barcoded libraries were normalized to 100pM using the Equalizer Kit (Thermo Fisher Scientific). Automated library preparation was performed using the OMR Assay-Chef Ready kit on the Ion Chef (Thermo Fisher Scientific). The normalized DNA and RNA libraries were diluted to optimized concentration and combined at a ratio of 80:20. Templating was performed on the Ion Chef according to manufacturer’s instructions (Thermo Fisher Scientific). The sample library was clonally amplified onto Ion Sphere Particles (ISP) by emulsion PCR with the Ion Chef System (Thermo Fisher Scientific) in line with the manufacturer’s instructions. Enriched ISPs were loaded onto 530 chips using an Ion 510 & Ion 520 & Ion 530 Kit-Chef. Eight and 12 samples were run in automated Chef and manual library preparation, respectively. Sequencing was performed on an Ion S5 Sequencer (Thermo Fisher Scientific). Sequence alignment to reference genome hg19 and base calling were performed using the Torrent Suite software (Version 5.10.0). Variant identification and annotation were performed using Ion Reporter software (Version 5.10) Coverage maps were generated using the Coverage Analysis plugin (Version 5.10.0).
The Ion Reporter default analysis parameter settings for Oncomine Myeloid Research workflow were used. In these settings, the minimum coverage requirement for the analysis is 20 for both single nucleotide variants (SNVs) and indels and 15 for hotspots; the minimum cut off variant allele fraction (VAF) is 2.5% for both SNVs and indels and 3% for hotspots; and the maximum strand bias tolerance is SNVs 0.9, indels 0.85, and hotspots 0.96. A separate custom DNA Myeloid filter chain was created in IR that allows for all possible pathogenic variants at a VAF ≤1%. The custom filter chain settings were: Variants type: SNV, Indel, MNV, CNV, Longdel, fusion,ExPR_Control,
Gene_Expression, RNAXonVariant, ProcControl, FLT3-ITD;
Variants effect: Effect in missense, nonframeshiftInsertion, nonframeshiftDeletion, nonframeshiftBlockSubstitution, nonsense, stoploss, frameshiftInsertion, frameshiftDeletion, frameshiftBlockSubstitution; 0≤VAF≤1%. The default RNA filter chain, oncomine_variant was used.
Results Comparison of gene panels of oncomine myeloid and illumina trusight As most of the DNA validation samples were previously characterized by the Illumina TMP, the gene sets of both panels were compared for gene content. The OMR panel contains RNA- and DNA-based gene sets that interrogate 23 hotspot genes, 17 full genes, 29 fusion driver genes, five expression genes, and five expression control genes. In comparison, the Illumina TMP interrogates DNA changes only and consists of 39 hotspot genes and 15 full genes. All hotspot genes covered by OMR are present in TMP; however, differences exist in the extent of exonic coverage of these hotspot genes. Thirteen of the 17 full genes covered by OMR are covered by
TMP. Of the non-overlapping genes, four genes were unique to OMR and 18 genes to ITM (Table 4). All recurrent gene rearrangements were unique to OMR.
Run Performance of OMR on ion S5 Library preparation was performed by both manual method and on the automated Ion Chef. Eight or 12 samples were loaded on the 530 chip after automated or manual library preparations, respectively. Quality metrics were assessed and compared between the two library prep methods, including total run reads, clonality, mean depth, uniformity, on-target reads, total mapped reads, and mean amplicon length. Both methods of library preparation showed identical performance in most of the parameters with the exception of higher mean depth for Chef library preparation because of the lower number of samples per chip on the Ion Chef (Table 5). Ninetyseven percent of bases within all targeted amplicons achieved greater than 500X coverage in every sample demonstrating adequate uniformity. The average coverage per amplicon was 3707 ±962 SD. For the chef library preparation, the average DNA on target was 97.7% ±0.2 SD. Some low coverage regions are observed consistently across all samples. Consistent coverage <500X was observed for three regions: BRAF chr7:140624442-140624670 (mean coverage 208, 96.5% below 500); BCOR chrX: 39923069-39923260 (mean coverage 185, 98.8% below 500); and PRPF8: chr17:1562643-1562846 (mean coverage 29, 100% below 500). A review of COSMIC (Catalogue of Somatic Mutations in Cancer) showed that these were non critical regions. There were no COSMIC regions under these BRAF and BCOR amplicon regions. The PRPF8 low coverage region had COSMIC mutations related to solid tumors, but no myeloid related mutations. However, three germline clinvar variants of uncertain clinical significance are
present. Other variants that may be clinically relevant that are known to be missed by OMR include: ASXL1 c.1934_1935insG; p.G646fs*12; CEBPA c.68_69insC; pH24fs*84 (homopolymer insertions); CEBPA c.1021_1022, ins882_898, inv920_1021; p.lle341Argfs*17 (variants occurring at edge of the amplicon); CEBPA c.dup914_1047 (134 nt duplication), p.Leu350Serfs*13; and FLT3-ITD mutations with lengths > 176bp.
Gene distribution by variant type From the 49 MDS, MPN, and AML patient samples, OMR detected 91unique DNA variants. Of these 91, 51 were SNVs and 40 were indels (≤ 5 bp=19, 5 to 40bp=15, and >40 bp=6; Figure 1). The sample selection in the validation study was non-random to include as many variants as possible, so the gene variant distribution does not necessarily reflect the natural distribution in myeloid disease (Figure 2, Figure 3). Among the 40 genes of the DNA panel, some variants were more commonly seen than others, with some genes having no variants detected in these validation samples. These genes include ABL1, GATA2, HRAS, MPL, MYD88, U2AF1, ETV6, IKZF1, PHF6, TP53, NF1, PRPF8, RB1, and SH2B3. However, variants of ABL1, MPL, MYD88, and U2F1 were detected in the SeraSeq DNA control.
Accuracy
DNA accuracy was assessed as the degree of agreement between the variants detected by OMR panel and the variants reported by the GetRM project of the reference Coriell cell line NA12878. There was a 100% concordance for all high-quality variants. Commercial Synthetic Seraseq DNA and RNA controls, and the IVS negative RNA control were also used to assess accuracy. All mutations in the Seraseq DNA control were detected except for 2 mutations in ASXL-1 and CEBPA giving a concordance of 91.3% (Figure 4; Table 6). All gene rearrangements in the seraseq RNA control were detected giving a concordance of 100% (Table 7). No gene fusions were detected in the IVS negative RNA control.
Analytical Sensitivity and Specificity Sensitivity and specificity was assessed based on the observed results of each target gene in comparison to reported results from the referral laboratory or the inhouse assay results. Hence each sample had true positive targets, when the variants detected were concordant with the reference or inhouse laboratory results; and true negatives when variants were undetected in concordance with the reference or inhouse laboratory results. Analytical sensitivity (PPA) is defined as the percentage of true positives divided by the sum of true positives and false negatives, whereas analytical specificity was defined as the percentage of false positives divided by the sum of false positives and true negatives. Fifty SNVs were detected, giving an analytical sensitivity of 100%. Of the indels, detection rates depended on the size. This ranged from 18/ 21(85.7%) of indels ≤5bp; 15/ 15 (100%) of indel 5 to 40bp and 6/ 6 (100%) of indel >40bp. Eleven of 12 CEBPA (91.7%) mutations were detected and 6/6(100%) FLT3-ITD were detected. The FLT3-ITD lengths varied from 21 to 54bp, with VAF ranging from 0.01 to 0.40. These
samples had been previously tested by PCR fragment analysis with capillary electrophoresis (PCR-CE) and TMP. Qualitative results (ie, positive or negative) showed good concordance between all three methods (OMR, TMP, and CE) at this fragment length and VAF range. The IR software also showed accurate calling ability. However, although the results from the NGS methods (OMR and TMP) showed a very close agreement in the values of VAF and ITD length, there was higher variation in comparison with the PCR-CE method. In total 89 of 92 mutations were detected (96.7%). Of these 89, 12 were new mutations, which were confirmed by orthogonal methods. Two (one SNVs and one indel) are as yet unconfirmed (Table 8). Analytical specificity was assessed with 28 negative samples. All samples had no reportable variants, giving an analytical specificity of 100%. To assess sensitivity and specificity for RNA fusion genes, 71 samples previously characterized by karyotyping, FISH, or molecular methods were analyzed for chromosomal alterations. Forty-three of 43 gene rearrangements were detected (100%). Two samples were excluded from analysis: One sample had a t(8;21) translocation that was below the limit of detection (LR>2.8 by quantitative real-time PCR); whereas the second failed QC metrics for RNA quality (Table 9). For analytical specificity, 26 samples with no chromosomal translocation and the IVS negative RNA control were analyzed. No gene fusions were detected, giving an analytical specificity of 100%.
Reproducibility To determine reproducibility of DNA variant detection, two samples were analyzed in triplicate over three runs and one sample in duplicate over three runs for the automated Ion Chef
library preparation. (The ion chef batch size is limited to eight samples restricting the number of samples that can be tested for a precision run.) The samples contained two, four, and five variants, respectively. Results showed 100% reproducibility for all variants. The VAF of all variants fell within a coefficient of variation (CV) of 1.4% to 7.7% of the mean (Table 10).
To assess reproducibility for gene fusion detection, two samples were examined in triplicate over three runs and one sample in duplicate over three runs. Reproducibility was 100% for the fusion genes tested. The RNA read counts fell within a range with CV of 11.9% to 37.3% of the mean (Table 11).
Limit of Detection The limit of detection (LOD) was assessed for SNVs and small indels using the Seraseq DNA control in which the variants and their allelic fraction have been characterized. VAF ranged from 4.5% to 18.3%. Serial dilutions of control in wild-type DNA was performed to yield a range of VAFs (Table 6). These dilutions were run in triplicate to determine reliable limits of detection. Reliability was determined by 100% detection in all replicates. The LOD is dependent on the specific variant but was generally detected near 5% for both SNVs and small indels. Of the 18 variants, CALR (52 bp deletion) LOD was 3.9% and FLT3-ITD LOD was 0.6%. Two out of the 20 variants showed higher LOD: U2AF1 c.101C>T, p.S34F at 10%, and CEBPA c.939_940insAAG at 15%.
To assess the LOD for RNA gene fusions, serial dilutions to 3-log reduction (LR) from original sample were performed on the Seraseq control with wild-type RNA. The LOD of gene fusion targets was possible down to 2-LR from initial baseline (Table 12). This NGS fusion panel gave accurate qualitative call and identified the primary breakpoints for downstream quantitative assays (Table 9).
Variant calling and annotation Some challenges of accuracy on automated Human Genome Variation Society (HGVS) nomenclature by Ion Reporter (IR) have been observed during validation, particularly with indels and complex variants. IR tends to misname horizontal complex variants such as deletioninsertions as SNVs, specifically when repetitive bases are present in the variants. Therefore, each filtered in variant needs to be manually viewed with the integrated genome viewer (IGV), and HGVS nomenclature and annotation need to be confirmed and documented if any change is made. The following variants are observed to be challenging for the IR software and should be reviewed with extra caution: WT1 c.1135delinsACCCT(called as WT1 c.1135G>A); WT1 c.1105_1107delinsGGGG; (called as WT1 c.1105C>G); RUNX1 c.254_257delACCCinsGAGAAGCT; (called as RUNX1 c.255C>G and RUNX1 c.257C>T); EZH2 c.1104delC; detectable but filtered out as sequencing strand error; and CEBPA c.246_247delCCinsG; detectable but masked due to low quality. Workflow Workflow was assessed in terms of steps, technologist hands on time, simplicity of analysis on the provided bioinformatics pipeline and turnaround time from library preparation (manual and
automated), templating on the Ion Chef, sequencing on the S5, analysis using Torrent Server Suite (TSS), and IR. The workflow using the Ion Chef and the S5 allowed high automation and reduced total hands on time to 45 minutes (includes sample preparation, quantitation, RT-PCR etc). Hands-on time for manual library preparation was seven hours. Library preparation runtime on the Chef was 7.5 hours for DNA, eight hours for RNA. Templating run time on the ion chef was 17 hours. Sequencing run time on the S5 was 4hours and 40 min. Analysis run time was 17 hours. Thus, a complete run time would take one day for library prep (consecutive DNA and RNA library prep on the chef), one day for templating, and one day for sequencing and analysis, with a total of three days before sign-out. Discussion Comprehensive characterization of myeloid neoplasms requires the assessment of variety of mutation types including SNVs, indels, and gene rearrangements. Current approaches combine a number of techniques to provide this comprehensive information for the management of these entities. Given the increasing number of targets and the need for rapid turnaround of results for prompt institution of treatment, a cost effective and rapid technology that can simultaneously interrogate all these mutation types is required. Here, we report the clinical validation and assessment of the OMR panel on the Ion Torrent S5 with the bioinformatics pipeline TSS and IR to meet the current demands of genomic characterization of myeloid neoplasm. The OMR on the Ion Torrent S5 was a highly accurate and highly reproducible panel for SNVs, small indels, and fusion genes. Both the bench work and bioinformatics analysis demonstrated high automation and simplicity of use and analysis. These features make this
panel and the sequencing system highly amenable for clinical testing of myeloid neoplasms. It is particularly useful for AML, which requires multiple testing methodologies including cytogenetic and molecular analysis in most cases for classification based on the WHO recommendation. The gene content is sufficiently comprehensive and includes the current recommended genes by the National Comprehensive Cancer Network and European Leukemia Network[16, 17]. This panel is also suitable and cost effective for the investigation of chronic myeloid neoplasms, including CEL, which requires assessment of gene rearrangements including PDGFRA, PDGFRB, FGFR1, PCM1-JAK2, BCR-ABL1, and molecular alterations in JAK2, KIT, and other CEL genes [15, 18]. The rapid output of three days is particularly advantageous for the acute TAT requirement of leukemias. The OMR also demonstrated very good sensitivity for fusion genes, with a detection limit of up to 2 log reduction (LR). This raises the possibility for quantitative monitoring for measurable residual disease (MRD), particularly in AML with recurrent cytogenetic abnormalities. However, issues of standardization, sample and run variability still need to be addressed. Normalization using total mapped reads and read counts of control genes may potentially mitigate sample and run variability. The impact of the complex metrics of the bioinformatics pipeline also requires investigation. It is worth noting that important clinical decision-making points/target of other established quantitative assays, eg,. t(9;22), t(8;21), inv(16), t(15;17) etc, occur at 3-4 LR [19-21], thus the OMR LOD of 2 LR may require improvement. A number of limitations were noted. Recently, the Association for Molecular Pathology (AMP) working group on leukemia published a list of minimum required genes to be included in a myeloid panel for chronic myeloid neoplasms[15]. Of the recommended core 34 genes, four
were absent from the OMR (BCORL1, PPM1D, RAD21, and SMC3). These genes have been identified with significant frequency in MDS (BCORL1-0.8%, RAD21-2%, and SMC3-2%), chronic myelomonocytic leukemia (BCORL1-1.8%) and clonal hematopoiesis of indeterminate significance (PPM1D-20%). An extensive review by the AMP working group attributed a low (RAD21, SMC3) to moderate (BCORL1, PPM1D) confidence score for their prognostic significance [22-24]. Supplementing the OMR with an additional panel may add to the cost of testing. The variant caller of the bioinformatics software, IR, has difficulty with horizontal complex variants in which deletions are replaced with stretches of repetitive bases. In addition, the OMR is unable to detect specific variants in ASXL1 and CEBPA in long homopolymer regions. This is a known limitation of the ion torrent chemistry for accurate variant calling of homopolymers above a stretch of eight or greater repeats, and in regions of high GC content. The detection of JAK2 Exon 12 variant c.1624_1629delAATGAA was accurately called after an upgrade of the software version. Some mutations, eg, EZH2 c.1104delC, though detected are deliberately blacklisted by the IR software due to the challenging chemistry. All such variants require careful review and visualization on a genome browser for more accurate calling. However, an inconsistent detection of the variant, ASXL1 c.1934_1935insG was observed in multiple known positive samples, despite manual review and visualization. We concluded that the basecalling software is not sufficiently reliable for these variants. Other potential work around need to be further explored, eg, the parallel use of other third-party commercial pipelines, one of which accurately detected this variant using the TMP panel. In conclusion, we report the clinical validation of the OMR panel on the Ion Torrent S5 and demonstrate it to be a highly accurate, reproducible with an efficient workflow that can meet the clinical demands for the genomic characterization of myeloid neoplasms.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
Arber, D.A., A. Orazi, R. Hasserjian, J. Thiele, M.J. Borowitz, M.M. Le Beau, C.D. Bloomfield, M. Cazzola, and J.W. Vardiman, The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood, 2016. 127(20): p. 2391-405. Cancer Genome Atlas Research, N., T.J. Ley, C. Miller, L. Ding, B.J. Raphael, A.J. Mungall, et al., Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med, 2013. 368(22): p. 2059-74. Papaemmanuil, E., M. Gerstung, L. Bullinger, V.I. Gaidzik, P. Paschka, N.D. Roberts, N.E. Potter, M. Heuser, F. Thol, N. Bolli, G. Gundem, P. Van Loo, I. Martincorena, P. Ganly, L. Mudie, S. McLaren, S. O'Meara, K. Raine, D.R. Jones, J.W. Teague, A.P. Butler, M.F. Greaves, A. Ganser, K. Dohner, R.F. Schlenk, H. Dohner, and P.J. Campbell, Genomic Classification and Prognosis in Acute Myeloid Leukemia. N Engl J Med, 2016. 374(23): p. 2209-2221. Klampfl, T., H. Gisslinger, A.S. Harutyunyan, H. Nivarthi, E. Rumi, J.D. Milosevic, N.C. Them, T. Berg, B. Gisslinger, D. Pietra, D. Chen, G.I. Vladimer, K. Bagienski, C. Milanesi, I.C. Casetti, E. Sant'Antonio, V. Ferretti, C. Elena, F. Schischlik, C. Cleary, M. Six, M. Schalling, A. Schonegger, C. Bock, L. Malcovati, C. Pascutto, G. Superti-Furga, M. Cazzola, and R. Kralovics, Somatic mutations of calreticulin in myeloproliferative neoplasms. N Engl J Med, 2013. 369(25): p. 2379-90. Kralovics, R., F. Passamonti, A.S. Buser, S.S. Teo, R. Tiedt, J.R. Passweg, A. Tichelli, M. Cazzola, and R.C. Skoda, A gain-of-function mutation of JAK2 in myeloproliferative disorders. N Engl J Med, 2005. 352(17): p. 1779-90. Nangalia, J., C.E. Massie, E.J. Baxter, F.L. Nice, G. Gundem, D.C. Wedge, et al., Somatic CALR mutations in myeloproliferative neoplasms with nonmutated JAK2. N Engl J Med, 2013. 369(25): p. 2391-2405. Bejar, R., K. Stevenson, O. Abdel-Wahab, N. Galili, B. Nilsson, G. Garcia-Manero, H. Kantarjian, A. Raza, R.L. Levine, D. Neuberg, and B.L. Ebert, Clinical effect of point mutations in myelodysplastic syndromes. N Engl J Med, 2011. 364(26): p. 2496-506. Papaemmanuil, E., M. Cazzola, J. Boultwood, L. Malcovati, P. Vyas, D. Bowen, et al., Somatic SF3B1 mutation in myelodysplasia with ring sideroblasts. N Engl J Med, 2011. 365(15): p. 138495. Papaemmanuil, E., M. Gerstung, L. Malcovati, S. Tauro, G. Gundem, P. Van Loo, et al., Clinical and biological implications of driver mutations in myelodysplastic syndromes. Blood, 2013. 122(22): p. 3616-27; quiz 3699. Reinig, E., F. Yang, E. Traer, R. Arora, S. Brown, R. Rattray, R. Braziel, G. Fan, R. Press, and J. Dunlap, Targeted Next-Generation Sequencing in Myelodysplastic Syndrome and Chronic Myelomonocytic Leukemia Aids Diagnosis in Challenging Cases and Identifies Frequent Spliceosome Mutations in Transformed Acute Myeloid Leukemia. Am J Clin Pathol, 2016. 145(4): p. 497-506. Au, C.H., A. Wa, D.N. Ho, T.L. Chan, and E.S. Ma, Clinical evaluation of panel testing by nextgeneration sequencing (NGS) for gene mutations in myeloid neoplasms. Diagn Pathol, 2016. 11: p. 11. Bartels, S., E. Schipper, B. Hasemeier, H. Kreipe, and U. Lehmann, Routine clinical mutation profiling using next generation sequencing and a customized gene panel improves diagnostic precision in myeloid neoplasms. Oncotarget, 2016. 7(21): p. 30084-93.
13.
14.
15.
16.
17.
18. 19.
20. 21.
22. 23.
24.
Kluk, M.J., R.C. Lindsley, J.C. Aster, N.I. Lindeman, D. Szeto, D. Hall, and F.C. Kuo, Validation and Implementation of a Custom Next-Generation Sequencing Clinical Assay for Hematologic Malignancies. J Mol Diagn, 2016. 18(4): p. 507-15. Thomas, M., M.A. Sukhai, T. Zhang, R. Dolatshahi, D. Harbi, S. Garg, M. Misyura, T. Pugh, T.L. Stockley, and S. Kamel-Reid, Integration of Technical, Bioinformatic, and Variant Assessment Approaches in the Validation of a Targeted Next-Generation Sequencing Panel for Myeloid Malignancies. Arch Pathol Lab Med, 2017. 141(6): p. 759-775. McClure, R.F., M.D. Ewalt, J. Crow, R.L. Temple-Smolkin, M. Pullambhatla, R. Sargent, and A.S. Kim, Clinical Significance of DNA Variants in Chronic Myeloid Neoplasms: A Report of the Association for Molecular Pathology. J Mol Diagn, 2018. 20(6): p. 717-737. Dohner, H., E. Estey, D. Grimwade, S. Amadori, F.R. Appelbaum, T. Buchner, H. Dombret, B.L. Ebert, P. Fenaux, R.A. Larson, R.L. Levine, F. Lo-Coco, T. Naoe, D. Niederwieser, G.J. Ossenkoppele, M. Sanz, J. Sierra, M.S. Tallman, H.F. Tien, A.H. Wei, B. Lowenberg, and C.D. Bloomfield, Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel. Blood, 2017. 129(4): p. 424-447. O'Donnell, M.R., M.S. Tallman, C.N. Abboud, J.K. Altman, F.R. Appelbaum, D.A. Arber, V. Bhatt, D. Bixby, W. Blum, S.E. Coutre, M. De Lima, A.T. Fathi, M. Fiorella, J.M. Foran, S.D. Gore, A.C. Hall, P. Kropf, J. Lancet, L.J. Maness, G. Marcucci, M.G. Martin, J.O. Moore, R. Olin, D. Peker, D.A. Pollyea, K. Pratz, F. Ravandi, P.J. Shami, R.M. Stone, S.A. Strickland, E.S. Wang, M. Wieduwilt, K. Gregory, and N. Ogba, Acute Myeloid Leukemia, Version 3.2017, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw, 2017. 15(7): p. 926-957. Klion, A.D., How I treat hypereosinophilic syndromes. Blood, 2015. 126(9): p. 1069-77. Chendamarai, E., P. Balasubramanian, B. George, A. Viswabandya, A. Abraham, R. Ahmed, A.A. Alex, S. Ganesan, K.M. Lakshmi, U. Sitaram, S.C. Nair, M. Chandy, N.B. Janet, V.M. Srivastava, A. Srivastava, and V. Mathews, Role of minimal residual disease monitoring in acute promyelocytic leukemia treated with arsenic trioxide in frontline therapy. Blood, 2012. 119(15): p. 3413-9. Ravandi, F., R.B. Walter, and S.D. Freeman, Evaluating measurable residual disease in acute myeloid leukemia. Blood Adv, 2018. 2(11): p. 1356-1366. Jourdan, E., N. Boissel, S. Chevret, E. Delabesse, A. Renneville, P. Cornillet, O. Blanchet, J.M. Cayuela, C. Recher, E. Raffoux, J. Delaunay, A. Pigneux, C.E. Bulabois, C. Berthon, C. Pautas, N. Vey, B. Lioure, X. Thomas, I. Luquet, C. Terre, P. Guardiola, M.C. Bene, C. Preudhomme, N. Ifrah, H. Dombret, and A.M.L.I. French, Prospective evaluation of gene mutations and minimal residual disease in patients with core binding factor acute myeloid leukemia. Blood, 2013. 121(12): p. 2213-23. Visconte, V., R.V. Tiu, and H.J. Rogers, Pathogenesis of myelodysplastic syndromes: an overview of molecular and non-molecular aspects of the disease. Blood Res, 2014. 49(4): p. 216-27. Damm, F., V. Chesnais, Y. Nagata, K. Yoshida, L. Scourzic, Y. Okuno, R. Itzykson, M. Sanada, Y. Shiraishi, V. Gelsi-Boyer, A. Renneville, S. Miyano, H. Mori, L.Y. Shih, S. Park, F. Dreyfus, A. Guerci-Bresler, E. Solary, C. Rose, S. Cheze, T. Prebet, N. Vey, M. Legentil, Y. Duffourd, S. de Botton, C. Preudhomme, D. Birnbaum, O.A. Bernard, S. Ogawa, M. Fontenay, and O. Kosmider, BCOR and BCORL1 mutations in myelodysplastic syndromes and related disorders. Blood, 2013. 122(18): p. 3169-77. Hsu, J.I., T. Dayaram, A. Tovy, E. De Braekeleer, M. Jeong, F. Wang, J. Zhang, T.P. Heffernan, S. Gera, J.J. Kovacs, J.R. Marszalek, C. Bristow, Y. Yan, G. Garcia-Manero, H. Kantarjian, G. Vassiliou, P.A. Futreal, L.A. Donehower, K. Takahashi, and M.A. Goodell, PPM1D Mutations Drive Clonal Hematopoiesis in Response to Cytotoxic Chemotherapy. Cell Stem Cell, 2018. 23(5): p. 700-713 e6.
Figure Legends Figure 1 DNA panel performance characteristics on clinical samples with variant AF>5%. Figure 2 Overall detected Oncomine Variant gene distributions by variant type. Figure 3 Gene fusion distribution in RNA Clinical Samples. Figure 4 SeraSeq DNA control accuracy.
Table 1 Genes in Oncomine Myeloid Research (OMR) Panel.
Hotspot genes (23) ABL1 KRAS
Full genes (17)
Fusion driver genes (29)
ASXL1 PRPF8
BRAF
MPL
BCOR RB1
CBL
MYD88
CALR RUNX1
ABL1 HMGA2 NUP214 ALK JAK2 PDGFRA BCL2 KMT2A PDGFRB BRAF ( MLL) RARA CCND1 MECOM RBM15 CREBBP MET RUNX1 EGFR MLLT10 TCF3 ETV6 MLLT3 TFE3 FGFR1 MYBL1 FGFR2 MYH11 FUS NTRK3
CSF3R NPM1 DNMT3A NRAS
CEBPA SH2B3 ETV6 STAG2
FLT3 PTPN11 GATA2 SETBP1 HRAS SF3B1 IDH1 SRSF2 IDH2 U2AF1 JAK2 WT1 KIT
EZH2 TET2 IKZF1 TP53 NF1 ZRSR2 PHF6
Expression Expression genes (5) control genes (5) BAALC EIF2B1 MECOM
FBXW2
MYC
PSMB2
SMC1A WT1
PUM1 TRIM27
Table 2 Sample selection: Samples were sourced from archival specimen previously analyzed by an inhouse single analyte or referral laboratory NGS diagnostic assay.
Sample Source
Number of samples
BM
PB
Comment
Myeloid NGS
67
58
9
Referred out samples
CALR
4
3
1
in house samples
FLIT3&NPM1 CEBPA
4 1
3 1
1 0
in house sample In house sample
KIT
1
1
0
in house sample
Total DNA Fusion (positive)
77 45
66 30
11 15
in house samples
Fusion (negative)
26
26
0
in house samples
Total RNA fusion
71
56
15
Total (DNA+RNA)
148
122
26
BM – Bone marrow; PB –Peripheral blood.
Table 3 Sample Selection for RNA Fusion Panel: Sample selected to include the most common gene fusions in AML with recurrent genetic abnormalities.
Type of Fusion
BCR-ABL1
PML-RARA
Type of Breakpoint
Number of Samples
t(9;22) b3a2
11
t(9;22) b2a2
6
t(9;22) e1a2 t(15;17) Long
2 2
t(15;17) Short
RUNX1-RUNX1T1
t(8;21)
MLLT3-MLL DEK-NUP214
t(9;11) t(6;9)
2 2(1 excluded from analysis) 3 1
inv(16) A
9
inv(16) D inv(16) E
2 1
inv(16)non-A/D/E
2 1(excluded from analysis) 1 45(2 excluded from analysis)
CBFB-MYH11
RPN1-EVI1(RPN1MECOM) FIP1L1-PDFRA Total
inv(3)
Total Number of Fusion type
19
4 2 3 1
14
1 1 45(2 excluded from analysis)
Gene fusions previously detected by karyotyping, molecular methods or FISH. Two samples failed QC parameters.
Table 4. Comparison of Genes and coverage between OMR and TMP shows differences in commercial NGS myeloid panels.
TMP gene Gene OMR gene coverage coverage ABL1 4-9 4-6 BRAF 1-4,6,8,11,15,17,18 15 CBL 8,9 8,9 CSF3R 14,17,18(NM_172313.2) 14-17 DNMT3A 11-23 full FLT3 8,11,13-16,20,23,24 14,15,20 GATA2 4,5 2-6 HRAS 2,3 2,3 IDH1 4 4 IDH2 4 4 JAK2 12-15 12,14 KIT 1,2,8-11,13,17 2,8-11,,13,17 KRAS 5-Feb 2,3 MPL 3,4,10,12 10 MYD88 2-5 3-5 NPM1 11 12 NRAS 4-Jan 2,3 PTPN11 3,12,13 3,13 SETBP1 4 4 (partial) SF3B1 14-21 13-16 SRSF2 1 1 U2AF1 2,6 2,6 WT1 7,9 7,9 ASXL1 full 12 BCOR full full CALR full 9 CEBPA full full ETV6 full full EZH2 full full
OMR gene Gene coverage IKZF1 full PHF6 full RUX1 full STAG2 full TET2 full TP53 full ZRSR2 full NF1 full PRPF8 full RB1 full SH2B3 full ATRX No BCORL1 No CBLB No CBLC No CDKN2A No CUX1 No FBXW7 No GATA1 No GNAS No JAK3 No KDM6A No MLL No NOTCH1 No PDGFRA No PTEN No RAD21 No SMC1A No SMC3 No
TMP gene coverage full full full full 3-11 2-11 full No No No No 8-10, 17-31 full 9,10 9,10 full full 9-11 2 8,9 13 full 5-8 26-28,34 12,14,18 5,7 full 2,11,16,17 10,13,19,23,25,28
Table 5. Overall Run Performance (Mean value): Comparison between the manual and automated library preparation. Library prep
Total Run Reads (M)
Clonalit y (%)
Mean depth
Uniform ity (%)
DNA on Target (%)
Total Mapped Reads on (DNA)
DNA Mean Amplico n length
Total Mapped Reads (RNA)
RNA Median Read Length
S5 Manual Library Prep
18.8
61.3
2652
99
97
1384314
227
168264
113
S5 Chef Library Prep
18.8
63.7
3707
98.3
97.7
1903608
224
232879
110
Mean
18.8
62.5
3180
99.1
97.4
1643961
225
200571
112
Table 6 SeraSeq DNA control Accuracy: Control sample (undiluted and in serial dilutions) was run in triplicate. VAF observed Gene ID Coding (Undilution) Expected Variants detected in all three 1 in 2 dilutions ABL1 c.944C>T 10% 10.5% BRAF c.1799T>A 10% 13.1% CALR c.1092_1143del52 5% 7.0% CBL -1 c.1139T>C 10% 18.3% CBL -2 c.1259G>A 5% 17.7% VAF expected
FLT3-1 FLT3 -2 FLT3 -3 JAK2 -1 MYD88
Dup chr13:28,608,25028,608,277 (hg19) c.2503G>T c.1759_1800dup c.1624_1629delAAT GAA c.794T>C
VAF observed (50% Dilution) Mean 4.2% 5.7% 3.9% 6.7% 6.6%
VAF observed (25% Dilution) Mean neg neg neg 4.8% neg
VAF observed (12.5% Dilution) Mean neg neg neg neg neg
10%
8.5%
3.1%
1.1%
0.8%
10% 5%
13.6% 5.2%
6.1% 1.7%
neg 0.8%
neg 0.6%
10%
11.0%
6.7%
neg
neg
10%
9.8%
5.0%
neg
neg
neg
neg
neg
neg
neg
neg
neg
neg
Expected Variants detected in 1 or 2 of the 1 in 2 dilutions SRSF2
c.284_307del24 _
5%
7.4%
NPM1
c.863_864insTCTG
5%
4.5%
SF3B1 -1
c.2098A>G
5%
6.5%
SF3B1-2
c.1998G>T
5%
6.8%
No Expected Variants detected in any dilutions CSF3R c.1853C>T 5% IDH1 c.394C>T 5% JAK2 -2 c.1849G>T 5% MPL c.1544G>T 5% U2AF1 c.101C>T 10% CEBPA VAF 15% , low sensitivity CEBPA -2 c.939_940insAAG 15% Variants undetectable by Oncomine Myeloid ASXL1-2 c.1934_1935insG 10% CEBPA-1 c.68_69insC 15%
4.2%/neg/ 4.7% 4.1%/neg/ neg neg/4.0%/ neg neg/3.5%/ neg
6.8% 5.7% 7.7% 7.9% 10.0%
neg neg neg neg neg
neg neg neg neg neg
neg neg neg neg neg
6.7
neg
neg
neg
-
-
The LOD and the number of positive replicates differ between variant type.
Table 7 SeraSeq RNA control Accuracy.
Fusion Partners
Oncomine Myeloid Research Assay Primary Detected Isoform
Read count
MYST3-CREBBP ETV6-ABL1 (transcript 1) ETV6-ABL1 (transcript 2) PCM1-JAK2 FIP1L1-PDGRFA TCF3-PBX1 BCR-ABL1 RUNX1-RUNX1T1
KAT6A-CREBBP.K17C2 ETV6-ABL1.E4A2 ETV6-ABL1.E5A2 PCM1-JAK2.P23J12.COSF1001 FIP1L1-PDGFRA.F11P12del45 TCF3-PBX1.T16P3.COSF1489 BCR-ABL1.B14A2.1 RUNX1-RUNX1T1.R3R3
2895 3637 5657 5042 3903 2513 3364 4129
PML-RARA
PML-RARA.P6del11ins133A3
3942
Table 8 Performance characteristics on DNA clinical samples with different mutation types and difficult target genes.
Variant Type
Known Variants
Known Variants Detected
PPA (%)
PPV (%)
SNV indel (≤5bp) indel (5-40bp) indel (>40bp) FLT3-ITD* CEBPA* Total
50 21 15 6 6 12 92
50 18 15 6 6 11 89
100.0 85.7 100.0 100.0 100.0 91.7 96.7
98.0 94.7 100.0 100.0 100.0 100.0 97.8
Features are characteristic of samples with VAF>5%, Total=SNV + indel ≤5bp + indel (5-40bp) + indel (>40bp). *included in SNV/indel count; PPV=TP/(TP+FP)x100, PPA =TP/(TP+FN)x100, SNV, single nucleotide variants; indel, insertion/deletion, TP, true positive; FP, false positive; FN, false negative; PPA, Positive Percentage Agreement; PPV, positive predictive value, VAF-variant allelic frequency.
Table 9. Performance characteristics on RNA clinical samples with gene fusions.
Fusion Type
Primary Breakpoints detected PML(6) - RARA(3)
Secondary Breakpoints detected -
Cases
PPA(%) PPV(%)
2
100
100
1
100
100
11
100
100
100
100
100
100
1
100
100
1
100
100
t(15:17) PML(3) - RARA(3) t(8;21)
RUNX1(3) - RUNX1T1(3) BCR(14)-ABL1(2)
t(9;22)
BCR(13) - ABL1(2)
BCR(13) - ABL1(2) BCR(1) - ABL1(2) CBFB(5) - MYH11(33) CBFB(5) - MYH11(29)
inv(16)
-
CBFB(5) - MYH11(28) CBFB(4) - MYH11(29) CBFB(5) - MYH11(32)
CBFB(4) MYH11(33)/ CBFB(4) MYH11(29) CBFB(4) MYH11(28) CBFB(4) MYH11(32) -
KMT2A(9) - MLLT3(6) t(9;11) t(6;9)
DEK(9) - NUP214(18)
FIP1L1-PDGFRA
FIP1L1(13)_PDGFRA(12)
-
Total
6 2 9 2 1 1 1 1
KMT2A(9) MLLT3(6) -
KMT2A(10) - MLLT3(6)
2
2
43
Detected breakpoints are shown. Some samples had an additional secondary breakpoint.TPTrue positive; FN-False Negative, FP-False positive; PPA -positive agreement value; PPV – positive predictive value.
Table 10. DNA Panel Precision: three samples each with multiple variants were run in triplicates or duplicates in three runs.
Variant Allele Frequency Sample
Variant
Run 1
Run 2
Run3
Within Run Mean
0.45
0.42
0.42
Within Run SD
0.01
0.00
0.01
Within Run CV%
3.1%
0.0%
1.7%
JAK2 c.1849G>T (p.Val617Phe) 0.35 Within Run Mean
0.36
0.35
0.38
Within Run SD
0.04
0.01
0.01
11.8%
2.0%
1.9%
0.32
0.35
0.37
Between Run Mean
SD
CV%
0.43
0.02
4.4%
0.36
0.02
4.2%
0.35
0.03
7.7%
0.40
0.02
4.3%
0.55
0.01
1.8%
0.76
0.01
1.4%
0.18
0.01
5.0%
0.44
0.01
1.5%
SF3B1 c.2098A>G (p.Lys700Glu) 0.43
1
Within Run CV% NPM1 c.863_864insTCTG 0.37 Within Run Mean Within Run SD
0.12
0.01
0.01
38.0%
1.6%
3.1%
NPM1 c.870G>A (p.Trp290Ter) 0.38 Within Run Mean
0.40
0.39
0.42
Within Run SD
0.05
0.03
0.02
11.4%
6.5%
4.8%
Within Run Mean
0.56
0.54
0.54
Within Run SD
0.02
0.02
0.02
Within Run CV%
3.6%
2.8%
3.8%
Within Run Mean
0.77
0.75
0.76
Within Run SD
0.03
0.01
0.01
Within Run CV%
3.3%
0.8%
0.8%
CSDE1 c.35G>T (p.Gly12Val) 0.18 Within Run Mean
0.17
0.17
0.19
Within Run SD
0.01
0.01
0.02
Within Run CV%
5.9%
6.7%
11.2%
0.43
0.44
0.45
Within Run CV%
Within Run CV% CEBPA c.1009A>T (p.Thr337Ser) 0.56 2
STAG2 c.1777C>T (p.Gin593Ter) 0.73
NPM1 c.867_868insGAAA (p.Trp290fs) 0.44 Within Run Mean 3
Within Run SD
0.01
0.01
0.03
Within Run CV%
2.7%
2.3%
6.8%
IDH1 c.395G>A (p.Arg132His)
0.35 Within Run Mean
0.38
0.35
0.37
Within Run SD
0.03
0.02
0.02
Within Run CV%
8.1%
5.7%
4.7%
IDH2 c.419G>A (p.Arg140Gln) 0.11 Within Run Mean
0.09
0.10
0.09
Within Run SD
0.01
0.02
0.01
Within Run CV%
6.5%
16.6%
6.7%
CSDE1 c.34G>A (p.Gly212Ser) 0.069 Within Run Mean
0.07
0.07
0.07
Within Run SD
0.01
0.02
0.01
14.3%
21.4%
14.8%
Within Run CV%
0.37
0.01
3.8%
0.10
0.00
2.7%
0.07
0.00
2.7%
Within run variation inVAF is shown in rows, between run variation in columns.
Table 11. RNA Fusion Panel Precision: three samples each with gene fusions were run in triplicates or duplicates in three runs.
Sample
Fusion Detected
Sample1 Within Run Mean Within Run SD Within Run CV%
CBFB(5) - MYH11(29)
Sample2 Within Run Mean Within Run SD With Run CV% Sample 3 Within Run Mean Within Run SD Within Run CV%
Read Count
Between Run
Run1
Run2
Run3
Mean
SD
CV%
10573 470 4.45%
16093 3359 20.87%
7689 299 3.89%
11451
4271
37.3%
RUNX1(3)RUNX1T1(3) 33866 3244 9.58%
44000 7605 17.28%
28963 2251 7.77%
35609
7669
21.5%
17306 2681 15.5%
20210 2464 12.2%
16075 1101 6.8%
17863
2123
11.9%
PML(3) - RARA(3)
Within run variation in read counts shown in rows, between run variation in columns.
Table 12. RNA Fusion Panel sensitivity: Serial dilution of RNA Control down to 2 log reduction.
Fusion Partners
Read Count undiluted
Read Count 10-1
Read Count 10-2
MYST3-CREBBP ETV6-ABL1 (transcript 1) ETV6-ABL1 (transcript 2) PCM1-JAK2 FIPIL1-PDGRFA
2895 3637 5657 5042 3903
321 341 552 405 541
35 43 59 48 76
TCF3-PBX1 BCR-ABL1 RUNX1-RUNX1T1 PML-RARA
2513 3364 4129 3942
425 366 334 410
40 46 40 71
S1525-1578(19)30409-X
DOI:
https://doi.org/10.1016/j.jmoldx.2019.10.002
Reference:
JMDI 845
To appear in:
The Journal of Molecular Diagnostics
Received Date: 5 June 2019 Revised Date:
11 September 2019
Accepted Date: 7 October 2019
Please cite this article as: Izevbaye, I, Liang, LY, Mather, C, El-Hallani, S, Maglantay JR, R, Saini L, Clinical Validation of a Myeloid Next-Generations Sequencing Panel for Single Nucleotide Variants, Indels, and Fusion Genes, The Journal of Molecular Diagnostics (2019), doi: https://doi.org/10.1016/ j.jmoldx.2019.10.002. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Copyright © 2019 Published by Elsevier Inc. on behalf of the American Society for Investigative Pathology and the Association for Molecular Pathology.
Clinical Validation of a Myeloid Next-Generations Sequencing Panel for Single Nucleotide Variants, Indels, and Fusion Genes Iyare Izevbaye,* Li Ying Liang,* Cheryl Mather,* Soufiane El-Hallani,* Remegio Maglantay JR,* and Lalit Saini† From the Department of Laboratory Medicine and Pathology,* University of Alberta Hospital, Edmonton, Alberta; and the Department of Medicine,† University of Alberta and Cross Cancer Institute, Edmonton, Alberta, Canada
Corresponding author: Iyare Izevbaye, Department of Laboratory Medicine and Pathology, 8440 112 Street NW, University of Alberta Hospital, Edmonton, AB, Canada T6G2B7, Tel: 780407-8025 Fax: 780-407-8599; Email: [email protected]
Short running title: Validation of a myeloid NGS panel
Funding: Supported by Marshall Eliuk Fund and University Hospital Foundation (SFR1404), University of Alberta, Alberta Public Laboratories, Edmonton; grant recipient: II.
Disclosures: I.I. has received honorarium for consultation/advisory role for Novartis, Pfizer, Roche, AstraZeneca, Bayer, Precision Rx-Dx.
Abstract Myeloid neoplasms are a heterogenous group of neoplasms including acute myeloid leukemia (AML), myeloproliferative neoplasms, myelodysplastic syndrome, and myeloproliferative neoplasms / myelodysplastic syndrome. Genetic abnormalities are used as diagnostic, prognostic, and predictive biomarkers in patients with these diseases. Here, we describe the clinical validation of the Oncomine Myeloid Research (OMR) next-generation sequencing panel that interrogates for 40 genes and 29 fusion genes commonly seen in myeloid neoplasms. Our validation set of 77 DNA samples included acute and chronic myeloid neoplasms with 91 single nucleotide variants and small indels. The 71 RNA samples from patients with AML included most of the AML-defining translocations. The OMR on the Ion Torrent S5 platform shows good performance in terms of depth of coverage, on-target reads, and uniformity. The panel achieved 91.3% and 100% concordance with reference DNA and RNA samples, with a clinical sensitivity and specificity of 96.7% and 100% for DNA, and 99.8% and 100% for RNA, respectively. Precision and reproducibility were 100% and the lower limit of detection was generally 5% VAF for DNA and 2-log reduction from initial value for RNA fusion genes. In conclusion, the OMR panel is a highly accurate and reproducible next-generation sequencing panel for the detection of common genetic alteration in myeloid neoplasms.
Introduction Next-generation sequencing (NGS) is an increasingly important diagnostic tool in the characterization of myeloid malignancies. For many years, World Health Organization (WHO) classification of acute myeloid leukemia (AML) has established the importance of recurrent cytogenetic and molecular features for diagnosis and prognosis [1]. According to the WHO scheme, class-defining cytogenetics and molecular features include recurrent gene rearrangements for t(15;17), t(8;21); inv(16)-t(16;16), t(9;22), t(9;11), t(6;9), inv(3)-t(3;3), t(1;22); and mutated NPM1, mutated CEBPA, and the provisional entity of mutated RUNX. FLT3 is also described as a poor prognostic marker in AML. With widespread application of genomic technology to the study of AML, many new molecular biomarkers have been shown to be prognostic and therapeutic targets. In a study of 200 adult AML cases, a total of 23 genes were significantly mutated. These genes could be placed in distinct functional categories including transcription factor fusions, tumor suppressor genes, DNA methylation-related genes, signaling genes, chromatin-modifying genes, myeloid transcription factor genes, cohesin-complex genes and spliceosome-complex genes. Functional relationships and genetic interplay between these genes were demonstrated by patterns of cooperation and mutual exclusivity seen in the mutations [2]. In another study of 1,540 patients, the authors argue for a genomic classification of AML based on driver mutation causality, an approach with clear implications for targeted therapy. Driver mutations were identified in 76 genes or genomic regions, with two or more drivers identified in 86% of patients. The authors suggested three genomic categories for AML including chromatin-spliceosome, TP53-aneuploidy, and IDH2R172 mutations. The genomic
categories and the presence or absence of other driver mutations were significant determinants of clinical outcomes [3]. Chronic myeloid neoplasms are a heterogenous group of myeloid neoplasms comprising myeloproliferative neoplasms (MPN), including chronic eosinophilic leukemia (CEL), myelodysplastic syndrome (MDS), and MPN/MDS. These myeloid entities are also diagnosed and classified based on molecular features. Chronic myelogenous leukemia (CML) is the classical molecularly defined neoplasm, in which the diagnostic t(9;22) translocation harbors comprehensive properties for diagnostic, predictive, monitoring, and resistance detection in the diagnosis and evolution of the disease. Non-BCR/ABL1 MPN also have defining genetic profiles with mutations in JAK2, CALR, and MPL [4-6]. Recurrent gene rearrangements defining the myeloid and lymphoid neoplasms with eosinophilia include abnormalities of PDGFRA, PDGFRB, FGFR and the newly identified PCM1-JAK2 translocation [1]. With the identification of new biomarkers and genetic targets, genomic characterization is becoming routine in the diagnosis and management of chronic myeloid neoplasms. NGS profiling has demonstrated utility for diagnosis, prognostic stratification, and therapeutic choice in myeloid neoplasms [7-10]. For example, SF3B1 mutations show correlation with MDS with ring sideroblasts9. Mutations in a number of genes such as TP53, EZH2, ETV6, RUNX1, and ASXL1 are independent prognostic predictors of reduced overall survival in MDS. NGS testing is frequently contributory in diagnostically difficult myeloid neoplasms by establishing clonality in borderline cases and differentiating neoplastic from reactive conditions [10]. This is particularly useful when morphologic and other routine molecular and cytogenetic tests are inconclusive.
Many different NGS myeloid panels including commercial and custom laboratorydeveloped tests (LDT) have been developed [11-14]. Consensus is still developing about the role of these biomarkers and which must be included in a myeloid panel. Recently, AMP released practice guidelines recommending a basic number of medically relevant genes that should be included in a myeloid panel [15]. Due to the multiplicity of genetic and cytogenetic alterations in myeloid malignancy and the need for prompt clinically required TAT, care must be taken in choosing an appropriate NGS assay that must meet these clinical requirements. Other important considerations include the local laboratory setting, the availability of technical manpower and process automation; and a sufficiently comprehensive spectrum of genes in the panel to cater for the variety of myeloid neoplasms to be managed in the clinical setting the laboratory serves. In this study, we report the clinical validation of the Oncomine Myeloid Research (OMR) panel on the Ion Torrent S5 (Thermo Fisher Scientific, San Francisco, CA). We studied the feasibility of a myeloid panel that is capable of providing comprehensive genomic information including recurrent cytogenetic analysis, SNVs, and small insertions-deletions (indels) for the diagnosis of a multiplicity of acute and chronic myeloid neoplasms. Examples of such entities having these diverse genetic aberrations include AML, MPN, MDS, and MDS/MPN. We assessed the performance of the OMR on the Ion Torrent S5 using samples previously characterized on the Illumina Trusight Myeloid panel (TMP; Illumina, San Diego, CA) and an Illumina custom panel designed by a referral laboratory, British Columbia Cancer Agency (BCCA).
Materials and Methods This study was reviewed and approved by the Health Research Ethics Board of Universities of Alberta and Calgary. OMR Panel content and genes/regions The OMR panel consists of RNA- and DNA-based gene panels with 526 DNA and 700 RNA amplicons. The average DNA amplicon size is 279 base pairs (bp) with an insert size of 227bp. The average RNA amplicon size is 143bp with an insert size of 99bp. The panel targets the complete exonic regions of 17 genes, exonic hot spots of 23 genes, 29 fusion genes, five expression genes, and five control genes (Table 1). Validation Samples DNA and RNA were extracted on the Qiasymphony instrument (Qiagen, Hilden, Germany) using the QiaSymphony kit from previously well-characterized peripheral blood (26) or bone marrow (122) samples. Nuclei acid was quantified by Nanodrop (Thermo Fisher Scientific) and Qubit HS Assay Kit (Thermo Fisher Scientific). Samples were run unpaired for DNA and RNA analysis. A total of 77 DNA samples were analyzed, comprising of 67 samples previously referred out for NGS analysis on the TMP panel or a custom illumina NGS myeloid panel; and an additional 10 DNA samples previously characterized for specific mutations (CALR, FLT3, KIT, and CEBPA) by in house methods. These additional samples were chosen to assess the capacity to sequence challenging genomic regions with high GC contents, egCEBPA and the limiting length of indels, egCALR and FLT3-ITD. The samples consisted of 49 samples with well characterized mutations and 28 negative samples with no pathogenic mutations. Benign, likely benign, variants of
undetermined significance, and low confidence variants were designated as no pathogenic mutation detected. The samples had a range of gene variants from 0 to 7. The variants included SNVs and indels of different sizes from ≤5, 5 to 40, >40bp (Table 2). The sample cohorts harbored variants in a spread of 26 unique genes characteristic of myeloid neoplasm. The diagnostic classes for DNA analysis included AML (42%), MPN (22%), MDS (6%), MDS/MPN (6%), clonal cytopenia of undetermined significance (CCUS; 4%), and others (20%).
A total of 71 RNA samples were analyzed, comprising of 45 samples with an abnormal fusion gene and 26 with normal or fusion negative karyotype. Two samples failed NGS analysis and were excluded from further assessment, giving a total of 43. The samples had been previously characterized by karyotyping and/or FISH. This selection covers seven of eight of the defined AML with recurrent genetic abnormalities and a FIP1L1-PDGFRA gene rearrangement (Table 3). The diagnostic classes for RNA analysis included AML (44.2%), acute promyelocytic leukemia (APL; 9.3%), acute lymphocytic leukemia (ALL; 4.6%), CML (39.5%), and CEL with PDGFRA gene rearrangement (2.3%). This selection covers all common AML with recurrent genetic abnormalities.
Commercial controls included in the analysis comprise Seraseq Myeloid Mutation DNA Mix (Seracare, Milford, MA; product number 0710-0408), having 23 clinically–relevant DNA mutations across 16 genes and the SeraSeq Myeloid Fusion RNA Mix (Seracare, product number 0710-0407) having nine RNA fusions. Well characterized Coriell wild-type DNA control (NA12878) (Coriell Institute of Medical Research, Camden, NJ) and IVS (Invivoscribe, San
Diego, CA) negative RNA control (IVS 0035) were used for preparation of dilutions for sensitivity testing for DNA and RNA panels, respectively, and for accuracy studies.
Library preparation, templating, and sequencing OMR panel library preparation was performed by manual and automated processes according to the manufacturer’s instructions (Thermo Fisher Scientific). Briefly, barcoded libraries were generated from 10 ng of sample gDNA and RNA. Targets were amplified using an RNA primer pool and two DNA primer pools using highly multiplexed PCR amplification. The amplicons were partially digested with Fupa enzyme followed by ligation of unique barcode adapters for each library. The barcoded libraries were normalized to 100pM using the Equalizer Kit (Thermo Fisher Scientific). Automated library preparation was performed using the OMR Assay-Chef Ready kit on the Ion Chef (Thermo Fisher Scientific). The normalized DNA and RNA libraries were diluted to optimized concentration and combined at a ratio of 80:20. Templating was performed on the Ion Chef according to manufacturer’s instructions (Thermo Fisher Scientific). The sample library was clonally amplified onto Ion Sphere Particles (ISP) by emulsion PCR with the Ion Chef System (Thermo Fisher Scientific) in line with the manufacturer’s instructions. Enriched ISPs were loaded onto 530 chips using an Ion 510 & Ion 520 & Ion 530 Kit-Chef. Eight and 12 samples were run in automated Chef and manual library preparation, respectively. Sequencing was performed on an Ion S5 Sequencer (Thermo Fisher Scientific). Sequence alignment to reference genome hg19 and base calling were performed using the Torrent Suite software (Version 5.10.0). Variant identification and annotation were performed using Ion Reporter software (Version 5.10) Coverage maps were generated using the Coverage Analysis plugin (Version 5.10.0).
The Ion Reporter default analysis parameter settings for Oncomine Myeloid Research workflow were used. In these settings, the minimum coverage requirement for the analysis is 20 for both single nucleotide variants (SNVs) and indels and 15 for hotspots; the minimum cut off variant allele fraction (VAF) is 2.5% for both SNVs and indels and 3% for hotspots; and the maximum strand bias tolerance is SNVs 0.9, indels 0.85, and hotspots 0.96. A separate custom DNA Myeloid filter chain was created in IR that allows for all possible pathogenic variants at a VAF ≤1%. The custom filter chain settings were: Variants type: SNV, Indel, MNV, CNV, Longdel, fusion,ExPR_Control,
Gene_Expression, RNAXonVariant, ProcControl, FLT3-ITD;
Variants effect: Effect in missense, nonframeshiftInsertion, nonframeshiftDeletion, nonframeshiftBlockSubstitution, nonsense, stoploss, frameshiftInsertion, frameshiftDeletion, frameshiftBlockSubstitution; 0≤VAF≤1%. The default RNA filter chain, oncomine_variant was used.
Results Comparison of gene panels of oncomine myeloid and illumina trusight As most of the DNA validation samples were previously characterized by the Illumina TMP, the gene sets of both panels were compared for gene content. The OMR panel contains RNA- and DNA-based gene sets that interrogate 23 hotspot genes, 17 full genes, 29 fusion driver genes, five expression genes, and five expression control genes. In comparison, the Illumina TMP interrogates DNA changes only and consists of 39 hotspot genes and 15 full genes. All hotspot genes covered by OMR are present in TMP; however, differences exist in the extent of exonic coverage of these hotspot genes. Thirteen of the 17 full genes covered by OMR are covered by
TMP. Of the non-overlapping genes, four genes were unique to OMR and 18 genes to ITM (Table 4). All recurrent gene rearrangements were unique to OMR.
Run Performance of OMR on ion S5 Library preparation was performed by both manual method and on the automated Ion Chef. Eight or 12 samples were loaded on the 530 chip after automated or manual library preparations, respectively. Quality metrics were assessed and compared between the two library prep methods, including total run reads, clonality, mean depth, uniformity, on-target reads, total mapped reads, and mean amplicon length. Both methods of library preparation showed identical performance in most of the parameters with the exception of higher mean depth for Chef library preparation because of the lower number of samples per chip on the Ion Chef (Table 5). Ninetyseven percent of bases within all targeted amplicons achieved greater than 500X coverage in every sample demonstrating adequate uniformity. The average coverage per amplicon was 3707 ±962 SD. For the chef library preparation, the average DNA on target was 97.7% ±0.2 SD. Some low coverage regions are observed consistently across all samples. Consistent coverage <500X was observed for three regions: BRAF chr7:140624442-140624670 (mean coverage 208, 96.5% below 500); BCOR chrX: 39923069-39923260 (mean coverage 185, 98.8% below 500); and PRPF8: chr17:1562643-1562846 (mean coverage 29, 100% below 500). A review of COSMIC (Catalogue of Somatic Mutations in Cancer) showed that these were non critical regions. There were no COSMIC regions under these BRAF and BCOR amplicon regions. The PRPF8 low coverage region had COSMIC mutations related to solid tumors, but no myeloid related mutations. However, three germline clinvar variants of uncertain clinical significance are
present. Other variants that may be clinically relevant that are known to be missed by OMR include: ASXL1 c.1934_1935insG; p.G646fs*12; CEBPA c.68_69insC; pH24fs*84 (homopolymer insertions); CEBPA c.1021_1022, ins882_898, inv920_1021; p.lle341Argfs*17 (variants occurring at edge of the amplicon); CEBPA c.dup914_1047 (134 nt duplication), p.Leu350Serfs*13; and FLT3-ITD mutations with lengths > 176bp.
Gene distribution by variant type From the 49 MDS, MPN, and AML patient samples, OMR detected 91unique DNA variants. Of these 91, 51 were SNVs and 40 were indels (≤ 5 bp=19, 5 to 40bp=15, and >40 bp=6; Figure 1). The sample selection in the validation study was non-random to include as many variants as possible, so the gene variant distribution does not necessarily reflect the natural distribution in myeloid disease (Figure 2, Figure 3). Among the 40 genes of the DNA panel, some variants were more commonly seen than others, with some genes having no variants detected in these validation samples. These genes include ABL1, GATA2, HRAS, MPL, MYD88, U2AF1, ETV6, IKZF1, PHF6, TP53, NF1, PRPF8, RB1, and SH2B3. However, variants of ABL1, MPL, MYD88, and U2F1 were detected in the SeraSeq DNA control.
Accuracy
DNA accuracy was assessed as the degree of agreement between the variants detected by OMR panel and the variants reported by the GetRM project of the reference Coriell cell line NA12878. There was a 100% concordance for all high-quality variants. Commercial Synthetic Seraseq DNA and RNA controls, and the IVS negative RNA control were also used to assess accuracy. All mutations in the Seraseq DNA control were detected except for 2 mutations in ASXL-1 and CEBPA giving a concordance of 91.3% (Figure 4; Table 6). All gene rearrangements in the seraseq RNA control were detected giving a concordance of 100% (Table 7). No gene fusions were detected in the IVS negative RNA control.
Analytical Sensitivity and Specificity Sensitivity and specificity was assessed based on the observed results of each target gene in comparison to reported results from the referral laboratory or the inhouse assay results. Hence each sample had true positive targets, when the variants detected were concordant with the reference or inhouse laboratory results; and true negatives when variants were undetected in concordance with the reference or inhouse laboratory results. Analytical sensitivity (PPA) is defined as the percentage of true positives divided by the sum of true positives and false negatives, whereas analytical specificity was defined as the percentage of false positives divided by the sum of false positives and true negatives. Fifty SNVs were detected, giving an analytical sensitivity of 100%. Of the indels, detection rates depended on the size. This ranged from 18/ 21(85.7%) of indels ≤5bp; 15/ 15 (100%) of indel 5 to 40bp and 6/ 6 (100%) of indel >40bp. Eleven of 12 CEBPA (91.7%) mutations were detected and 6/6(100%) FLT3-ITD were detected. The FLT3-ITD lengths varied from 21 to 54bp, with VAF ranging from 0.01 to 0.40. These
samples had been previously tested by PCR fragment analysis with capillary electrophoresis (PCR-CE) and TMP. Qualitative results (ie, positive or negative) showed good concordance between all three methods (OMR, TMP, and CE) at this fragment length and VAF range. The IR software also showed accurate calling ability. However, although the results from the NGS methods (OMR and TMP) showed a very close agreement in the values of VAF and ITD length, there was higher variation in comparison with the PCR-CE method. In total 89 of 92 mutations were detected (96.7%). Of these 89, 12 were new mutations, which were confirmed by orthogonal methods. Two (one SNVs and one indel) are as yet unconfirmed (Table 8). Analytical specificity was assessed with 28 negative samples. All samples had no reportable variants, giving an analytical specificity of 100%. To assess sensitivity and specificity for RNA fusion genes, 71 samples previously characterized by karyotyping, FISH, or molecular methods were analyzed for chromosomal alterations. Forty-three of 43 gene rearrangements were detected (100%). Two samples were excluded from analysis: One sample had a t(8;21) translocation that was below the limit of detection (LR>2.8 by quantitative real-time PCR); whereas the second failed QC metrics for RNA quality (Table 9). For analytical specificity, 26 samples with no chromosomal translocation and the IVS negative RNA control were analyzed. No gene fusions were detected, giving an analytical specificity of 100%.
Reproducibility To determine reproducibility of DNA variant detection, two samples were analyzed in triplicate over three runs and one sample in duplicate over three runs for the automated Ion Chef
library preparation. (The ion chef batch size is limited to eight samples restricting the number of samples that can be tested for a precision run.) The samples contained two, four, and five variants, respectively. Results showed 100% reproducibility for all variants. The VAF of all variants fell within a coefficient of variation (CV) of 1.4% to 7.7% of the mean (Table 10).
To assess reproducibility for gene fusion detection, two samples were examined in triplicate over three runs and one sample in duplicate over three runs. Reproducibility was 100% for the fusion genes tested. The RNA read counts fell within a range with CV of 11.9% to 37.3% of the mean (Table 11).
Limit of Detection The limit of detection (LOD) was assessed for SNVs and small indels using the Seraseq DNA control in which the variants and their allelic fraction have been characterized. VAF ranged from 4.5% to 18.3%. Serial dilutions of control in wild-type DNA was performed to yield a range of VAFs (Table 6). These dilutions were run in triplicate to determine reliable limits of detection. Reliability was determined by 100% detection in all replicates. The LOD is dependent on the specific variant but was generally detected near 5% for both SNVs and small indels. Of the 18 variants, CALR (52 bp deletion) LOD was 3.9% and FLT3-ITD LOD was 0.6%. Two out of the 20 variants showed higher LOD: U2AF1 c.101C>T, p.S34F at 10%, and CEBPA c.939_940insAAG at 15%.
To assess the LOD for RNA gene fusions, serial dilutions to 3-log reduction (LR) from original sample were performed on the Seraseq control with wild-type RNA. The LOD of gene fusion targets was possible down to 2-LR from initial baseline (Table 12). This NGS fusion panel gave accurate qualitative call and identified the primary breakpoints for downstream quantitative assays (Table 9).
Variant calling and annotation Some challenges of accuracy on automated Human Genome Variation Society (HGVS) nomenclature by Ion Reporter (IR) have been observed during validation, particularly with indels and complex variants. IR tends to misname horizontal complex variants such as deletioninsertions as SNVs, specifically when repetitive bases are present in the variants. Therefore, each filtered in variant needs to be manually viewed with the integrated genome viewer (IGV), and HGVS nomenclature and annotation need to be confirmed and documented if any change is made. The following variants are observed to be challenging for the IR software and should be reviewed with extra caution: WT1 c.1135delinsACCCT(called as WT1 c.1135G>A); WT1 c.1105_1107delinsGGGG; (called as WT1 c.1105C>G); RUNX1 c.254_257delACCCinsGAGAAGCT; (called as RUNX1 c.255C>G and RUNX1 c.257C>T); EZH2 c.1104delC; detectable but filtered out as sequencing strand error; and CEBPA c.246_247delCCinsG; detectable but masked due to low quality. Workflow Workflow was assessed in terms of steps, technologist hands on time, simplicity of analysis on the provided bioinformatics pipeline and turnaround time from library preparation (manual and
automated), templating on the Ion Chef, sequencing on the S5, analysis using Torrent Server Suite (TSS), and IR. The workflow using the Ion Chef and the S5 allowed high automation and reduced total hands on time to 45 minutes (includes sample preparation, quantitation, RT-PCR etc). Hands-on time for manual library preparation was seven hours. Library preparation runtime on the Chef was 7.5 hours for DNA, eight hours for RNA. Templating run time on the ion chef was 17 hours. Sequencing run time on the S5 was 4hours and 40 min. Analysis run time was 17 hours. Thus, a complete run time would take one day for library prep (consecutive DNA and RNA library prep on the chef), one day for templating, and one day for sequencing and analysis, with a total of three days before sign-out. Discussion Comprehensive characterization of myeloid neoplasms requires the assessment of variety of mutation types including SNVs, indels, and gene rearrangements. Current approaches combine a number of techniques to provide this comprehensive information for the management of these entities. Given the increasing number of targets and the need for rapid turnaround of results for prompt institution of treatment, a cost effective and rapid technology that can simultaneously interrogate all these mutation types is required. Here, we report the clinical validation and assessment of the OMR panel on the Ion Torrent S5 with the bioinformatics pipeline TSS and IR to meet the current demands of genomic characterization of myeloid neoplasm. The OMR on the Ion Torrent S5 was a highly accurate and highly reproducible panel for SNVs, small indels, and fusion genes. Both the bench work and bioinformatics analysis demonstrated high automation and simplicity of use and analysis. These features make this
panel and the sequencing system highly amenable for clinical testing of myeloid neoplasms. It is particularly useful for AML, which requires multiple testing methodologies including cytogenetic and molecular analysis in most cases for classification based on the WHO recommendation. The gene content is sufficiently comprehensive and includes the current recommended genes by the National Comprehensive Cancer Network and European Leukemia Network[16, 17]. This panel is also suitable and cost effective for the investigation of chronic myeloid neoplasms, including CEL, which requires assessment of gene rearrangements including PDGFRA, PDGFRB, FGFR1, PCM1-JAK2, BCR-ABL1, and molecular alterations in JAK2, KIT, and other CEL genes [15, 18]. The rapid output of three days is particularly advantageous for the acute TAT requirement of leukemias. The OMR also demonstrated very good sensitivity for fusion genes, with a detection limit of up to 2 log reduction (LR). This raises the possibility for quantitative monitoring for measurable residual disease (MRD), particularly in AML with recurrent cytogenetic abnormalities. However, issues of standardization, sample and run variability still need to be addressed. Normalization using total mapped reads and read counts of control genes may potentially mitigate sample and run variability. The impact of the complex metrics of the bioinformatics pipeline also requires investigation. It is worth noting that important clinical decision-making points/target of other established quantitative assays, eg,. t(9;22), t(8;21), inv(16), t(15;17) etc, occur at 3-4 LR [19-21], thus the OMR LOD of 2 LR may require improvement. A number of limitations were noted. Recently, the Association for Molecular Pathology (AMP) working group on leukemia published a list of minimum required genes to be included in a myeloid panel for chronic myeloid neoplasms[15]. Of the recommended core 34 genes, four
were absent from the OMR (BCORL1, PPM1D, RAD21, and SMC3). These genes have been identified with significant frequency in MDS (BCORL1-0.8%, RAD21-2%, and SMC3-2%), chronic myelomonocytic leukemia (BCORL1-1.8%) and clonal hematopoiesis of indeterminate significance (PPM1D-20%). An extensive review by the AMP working group attributed a low (RAD21, SMC3) to moderate (BCORL1, PPM1D) confidence score for their prognostic significance [22-24]. Supplementing the OMR with an additional panel may add to the cost of testing. The variant caller of the bioinformatics software, IR, has difficulty with horizontal complex variants in which deletions are replaced with stretches of repetitive bases. In addition, the OMR is unable to detect specific variants in ASXL1 and CEBPA in long homopolymer regions. This is a known limitation of the ion torrent chemistry for accurate variant calling of homopolymers above a stretch of eight or greater repeats, and in regions of high GC content. The detection of JAK2 Exon 12 variant c.1624_1629delAATGAA was accurately called after an upgrade of the software version. Some mutations, eg, EZH2 c.1104delC, though detected are deliberately blacklisted by the IR software due to the challenging chemistry. All such variants require careful review and visualization on a genome browser for more accurate calling. However, an inconsistent detection of the variant, ASXL1 c.1934_1935insG was observed in multiple known positive samples, despite manual review and visualization. We concluded that the basecalling software is not sufficiently reliable for these variants. Other potential work around need to be further explored, eg, the parallel use of other third-party commercial pipelines, one of which accurately detected this variant using the TMP panel. In conclusion, we report the clinical validation of the OMR panel on the Ion Torrent S5 and demonstrate it to be a highly accurate, reproducible with an efficient workflow that can meet the clinical demands for the genomic characterization of myeloid neoplasms.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
Arber, D.A., A. Orazi, R. Hasserjian, J. Thiele, M.J. Borowitz, M.M. Le Beau, C.D. Bloomfield, M. Cazzola, and J.W. Vardiman, The 2016 revision to the World Health Organization classification of myeloid neoplasms and acute leukemia. Blood, 2016. 127(20): p. 2391-405. Cancer Genome Atlas Research, N., T.J. Ley, C. Miller, L. Ding, B.J. Raphael, A.J. Mungall, et al., Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med, 2013. 368(22): p. 2059-74. Papaemmanuil, E., M. Gerstung, L. Bullinger, V.I. Gaidzik, P. Paschka, N.D. Roberts, N.E. Potter, M. Heuser, F. Thol, N. Bolli, G. Gundem, P. Van Loo, I. Martincorena, P. Ganly, L. Mudie, S. McLaren, S. O'Meara, K. Raine, D.R. Jones, J.W. Teague, A.P. Butler, M.F. Greaves, A. Ganser, K. Dohner, R.F. Schlenk, H. Dohner, and P.J. Campbell, Genomic Classification and Prognosis in Acute Myeloid Leukemia. N Engl J Med, 2016. 374(23): p. 2209-2221. Klampfl, T., H. Gisslinger, A.S. Harutyunyan, H. Nivarthi, E. Rumi, J.D. Milosevic, N.C. Them, T. Berg, B. Gisslinger, D. Pietra, D. Chen, G.I. Vladimer, K. Bagienski, C. Milanesi, I.C. Casetti, E. Sant'Antonio, V. Ferretti, C. Elena, F. Schischlik, C. Cleary, M. Six, M. Schalling, A. Schonegger, C. Bock, L. Malcovati, C. Pascutto, G. Superti-Furga, M. Cazzola, and R. Kralovics, Somatic mutations of calreticulin in myeloproliferative neoplasms. N Engl J Med, 2013. 369(25): p. 2379-90. Kralovics, R., F. Passamonti, A.S. Buser, S.S. Teo, R. Tiedt, J.R. Passweg, A. Tichelli, M. Cazzola, and R.C. Skoda, A gain-of-function mutation of JAK2 in myeloproliferative disorders. N Engl J Med, 2005. 352(17): p. 1779-90. Nangalia, J., C.E. Massie, E.J. Baxter, F.L. Nice, G. Gundem, D.C. Wedge, et al., Somatic CALR mutations in myeloproliferative neoplasms with nonmutated JAK2. N Engl J Med, 2013. 369(25): p. 2391-2405. Bejar, R., K. Stevenson, O. Abdel-Wahab, N. Galili, B. Nilsson, G. Garcia-Manero, H. Kantarjian, A. Raza, R.L. Levine, D. Neuberg, and B.L. Ebert, Clinical effect of point mutations in myelodysplastic syndromes. N Engl J Med, 2011. 364(26): p. 2496-506. Papaemmanuil, E., M. Cazzola, J. Boultwood, L. Malcovati, P. Vyas, D. Bowen, et al., Somatic SF3B1 mutation in myelodysplasia with ring sideroblasts. N Engl J Med, 2011. 365(15): p. 138495. Papaemmanuil, E., M. Gerstung, L. Malcovati, S. Tauro, G. Gundem, P. Van Loo, et al., Clinical and biological implications of driver mutations in myelodysplastic syndromes. Blood, 2013. 122(22): p. 3616-27; quiz 3699. Reinig, E., F. Yang, E. Traer, R. Arora, S. Brown, R. Rattray, R. Braziel, G. Fan, R. Press, and J. Dunlap, Targeted Next-Generation Sequencing in Myelodysplastic Syndrome and Chronic Myelomonocytic Leukemia Aids Diagnosis in Challenging Cases and Identifies Frequent Spliceosome Mutations in Transformed Acute Myeloid Leukemia. Am J Clin Pathol, 2016. 145(4): p. 497-506. Au, C.H., A. Wa, D.N. Ho, T.L. Chan, and E.S. Ma, Clinical evaluation of panel testing by nextgeneration sequencing (NGS) for gene mutations in myeloid neoplasms. Diagn Pathol, 2016. 11: p. 11. Bartels, S., E. Schipper, B. Hasemeier, H. Kreipe, and U. Lehmann, Routine clinical mutation profiling using next generation sequencing and a customized gene panel improves diagnostic precision in myeloid neoplasms. Oncotarget, 2016. 7(21): p. 30084-93.
13.
14.
15.
16.
17.
18. 19.
20. 21.
22. 23.
24.
Kluk, M.J., R.C. Lindsley, J.C. Aster, N.I. Lindeman, D. Szeto, D. Hall, and F.C. Kuo, Validation and Implementation of a Custom Next-Generation Sequencing Clinical Assay for Hematologic Malignancies. J Mol Diagn, 2016. 18(4): p. 507-15. Thomas, M., M.A. Sukhai, T. Zhang, R. Dolatshahi, D. Harbi, S. Garg, M. Misyura, T. Pugh, T.L. Stockley, and S. Kamel-Reid, Integration of Technical, Bioinformatic, and Variant Assessment Approaches in the Validation of a Targeted Next-Generation Sequencing Panel for Myeloid Malignancies. Arch Pathol Lab Med, 2017. 141(6): p. 759-775. McClure, R.F., M.D. Ewalt, J. Crow, R.L. Temple-Smolkin, M. Pullambhatla, R. Sargent, and A.S. Kim, Clinical Significance of DNA Variants in Chronic Myeloid Neoplasms: A Report of the Association for Molecular Pathology. J Mol Diagn, 2018. 20(6): p. 717-737. Dohner, H., E. Estey, D. Grimwade, S. Amadori, F.R. Appelbaum, T. Buchner, H. Dombret, B.L. Ebert, P. Fenaux, R.A. Larson, R.L. Levine, F. Lo-Coco, T. Naoe, D. Niederwieser, G.J. Ossenkoppele, M. Sanz, J. Sierra, M.S. Tallman, H.F. Tien, A.H. Wei, B. Lowenberg, and C.D. Bloomfield, Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel. Blood, 2017. 129(4): p. 424-447. O'Donnell, M.R., M.S. Tallman, C.N. Abboud, J.K. Altman, F.R. Appelbaum, D.A. Arber, V. Bhatt, D. Bixby, W. Blum, S.E. Coutre, M. De Lima, A.T. Fathi, M. Fiorella, J.M. Foran, S.D. Gore, A.C. Hall, P. Kropf, J. Lancet, L.J. Maness, G. Marcucci, M.G. Martin, J.O. Moore, R. Olin, D. Peker, D.A. Pollyea, K. Pratz, F. Ravandi, P.J. Shami, R.M. Stone, S.A. Strickland, E.S. Wang, M. Wieduwilt, K. Gregory, and N. Ogba, Acute Myeloid Leukemia, Version 3.2017, NCCN Clinical Practice Guidelines in Oncology. J Natl Compr Canc Netw, 2017. 15(7): p. 926-957. Klion, A.D., How I treat hypereosinophilic syndromes. Blood, 2015. 126(9): p. 1069-77. Chendamarai, E., P. Balasubramanian, B. George, A. Viswabandya, A. Abraham, R. Ahmed, A.A. Alex, S. Ganesan, K.M. Lakshmi, U. Sitaram, S.C. Nair, M. Chandy, N.B. Janet, V.M. Srivastava, A. Srivastava, and V. Mathews, Role of minimal residual disease monitoring in acute promyelocytic leukemia treated with arsenic trioxide in frontline therapy. Blood, 2012. 119(15): p. 3413-9. Ravandi, F., R.B. Walter, and S.D. Freeman, Evaluating measurable residual disease in acute myeloid leukemia. Blood Adv, 2018. 2(11): p. 1356-1366. Jourdan, E., N. Boissel, S. Chevret, E. Delabesse, A. Renneville, P. Cornillet, O. Blanchet, J.M. Cayuela, C. Recher, E. Raffoux, J. Delaunay, A. Pigneux, C.E. Bulabois, C. Berthon, C. Pautas, N. Vey, B. Lioure, X. Thomas, I. Luquet, C. Terre, P. Guardiola, M.C. Bene, C. Preudhomme, N. Ifrah, H. Dombret, and A.M.L.I. French, Prospective evaluation of gene mutations and minimal residual disease in patients with core binding factor acute myeloid leukemia. Blood, 2013. 121(12): p. 2213-23. Visconte, V., R.V. Tiu, and H.J. Rogers, Pathogenesis of myelodysplastic syndromes: an overview of molecular and non-molecular aspects of the disease. Blood Res, 2014. 49(4): p. 216-27. Damm, F., V. Chesnais, Y. Nagata, K. Yoshida, L. Scourzic, Y. Okuno, R. Itzykson, M. Sanada, Y. Shiraishi, V. Gelsi-Boyer, A. Renneville, S. Miyano, H. Mori, L.Y. Shih, S. Park, F. Dreyfus, A. Guerci-Bresler, E. Solary, C. Rose, S. Cheze, T. Prebet, N. Vey, M. Legentil, Y. Duffourd, S. de Botton, C. Preudhomme, D. Birnbaum, O.A. Bernard, S. Ogawa, M. Fontenay, and O. Kosmider, BCOR and BCORL1 mutations in myelodysplastic syndromes and related disorders. Blood, 2013. 122(18): p. 3169-77. Hsu, J.I., T. Dayaram, A. Tovy, E. De Braekeleer, M. Jeong, F. Wang, J. Zhang, T.P. Heffernan, S. Gera, J.J. Kovacs, J.R. Marszalek, C. Bristow, Y. Yan, G. Garcia-Manero, H. Kantarjian, G. Vassiliou, P.A. Futreal, L.A. Donehower, K. Takahashi, and M.A. Goodell, PPM1D Mutations Drive Clonal Hematopoiesis in Response to Cytotoxic Chemotherapy. Cell Stem Cell, 2018. 23(5): p. 700-713 e6.
Figure Legends Figure 1 DNA panel performance characteristics on clinical samples with variant AF>5%. Figure 2 Overall detected Oncomine Variant gene distributions by variant type. Figure 3 Gene fusion distribution in RNA Clinical Samples. Figure 4 SeraSeq DNA control accuracy.
Table 1 Genes in Oncomine Myeloid Research (OMR) Panel.
Hotspot genes (23) ABL1 KRAS
Full genes (17)
Fusion driver genes (29)
ASXL1 PRPF8
BRAF
MPL
BCOR RB1
CBL
MYD88
CALR RUNX1
ABL1 HMGA2 NUP214 ALK JAK2 PDGFRA BCL2 KMT2A PDGFRB BRAF ( MLL) RARA CCND1 MECOM RBM15 CREBBP MET RUNX1 EGFR MLLT10 TCF3 ETV6 MLLT3 TFE3 FGFR1 MYBL1 FGFR2 MYH11 FUS NTRK3
CSF3R NPM1 DNMT3A NRAS
CEBPA SH2B3 ETV6 STAG2
FLT3 PTPN11 GATA2 SETBP1 HRAS SF3B1 IDH1 SRSF2 IDH2 U2AF1 JAK2 WT1 KIT
EZH2 TET2 IKZF1 TP53 NF1 ZRSR2 PHF6
Expression Expression genes (5) control genes (5) BAALC EIF2B1 MECOM
FBXW2
MYC
PSMB2
SMC1A WT1
PUM1 TRIM27
Table 2 Sample selection: Samples were sourced from archival specimen previously analyzed by an inhouse single analyte or referral laboratory NGS diagnostic assay.
Sample Source
Number of samples
BM
PB
Comment
Myeloid NGS
67
58
9
Referred out samples
CALR
4
3
1
in house samples
FLIT3&NPM1 CEBPA
4 1
3 1
1 0
in house sample In house sample
KIT
1
1
0
in house sample
Total DNA Fusion (positive)
77 45
66 30
11 15
in house samples
Fusion (negative)
26
26
0
in house samples
Total RNA fusion
71
56
15
Total (DNA+RNA)
148
122
26
BM – Bone marrow; PB –Peripheral blood.
Table 3 Sample Selection for RNA Fusion Panel: Sample selected to include the most common gene fusions in AML with recurrent genetic abnormalities.
Type of Fusion
BCR-ABL1
PML-RARA
Type of Breakpoint
Number of Samples
t(9;22) b3a2
11
t(9;22) b2a2
6
t(9;22) e1a2 t(15;17) Long
2 2
t(15;17) Short
RUNX1-RUNX1T1
t(8;21)
MLLT3-MLL DEK-NUP214
t(9;11) t(6;9)
2 2(1 excluded from analysis) 3 1
inv(16) A
9
inv(16) D inv(16) E
2 1
inv(16)non-A/D/E
2 1(excluded from analysis) 1 45(2 excluded from analysis)
CBFB-MYH11
RPN1-EVI1(RPN1MECOM) FIP1L1-PDFRA Total
inv(3)
Total Number of Fusion type
19
4 2 3 1
14
1 1 45(2 excluded from analysis)
Gene fusions previously detected by karyotyping, molecular methods or FISH. Two samples failed QC parameters.
Table 4. Comparison of Genes and coverage between OMR and TMP shows differences in commercial NGS myeloid panels.
TMP gene Gene OMR gene coverage coverage ABL1 4-9 4-6 BRAF 1-4,6,8,11,15,17,18 15 CBL 8,9 8,9 CSF3R 14,17,18(NM_172313.2) 14-17 DNMT3A 11-23 full FLT3 8,11,13-16,20,23,24 14,15,20 GATA2 4,5 2-6 HRAS 2,3 2,3 IDH1 4 4 IDH2 4 4 JAK2 12-15 12,14 KIT 1,2,8-11,13,17 2,8-11,,13,17 KRAS 5-Feb 2,3 MPL 3,4,10,12 10 MYD88 2-5 3-5 NPM1 11 12 NRAS 4-Jan 2,3 PTPN11 3,12,13 3,13 SETBP1 4 4 (partial) SF3B1 14-21 13-16 SRSF2 1 1 U2AF1 2,6 2,6 WT1 7,9 7,9 ASXL1 full 12 BCOR full full CALR full 9 CEBPA full full ETV6 full full EZH2 full full
OMR gene Gene coverage IKZF1 full PHF6 full RUX1 full STAG2 full TET2 full TP53 full ZRSR2 full NF1 full PRPF8 full RB1 full SH2B3 full ATRX No BCORL1 No CBLB No CBLC No CDKN2A No CUX1 No FBXW7 No GATA1 No GNAS No JAK3 No KDM6A No MLL No NOTCH1 No PDGFRA No PTEN No RAD21 No SMC1A No SMC3 No
TMP gene coverage full full full full 3-11 2-11 full No No No No 8-10, 17-31 full 9,10 9,10 full full 9-11 2 8,9 13 full 5-8 26-28,34 12,14,18 5,7 full 2,11,16,17 10,13,19,23,25,28
Table 5. Overall Run Performance (Mean value): Comparison between the manual and automated library preparation. Library prep
Total Run Reads (M)
Clonalit y (%)
Mean depth
Uniform ity (%)
DNA on Target (%)
Total Mapped Reads on (DNA)
DNA Mean Amplico n length
Total Mapped Reads (RNA)
RNA Median Read Length
S5 Manual Library Prep
18.8
61.3
2652
99
97
1384314
227
168264
113
S5 Chef Library Prep
18.8
63.7
3707
98.3
97.7
1903608
224
232879
110
Mean
18.8
62.5
3180
99.1
97.4
1643961
225
200571
112
Table 6 SeraSeq DNA control Accuracy: Control sample (undiluted and in serial dilutions) was run in triplicate. VAF observed Gene ID Coding (Undilution) Expected Variants detected in all three 1 in 2 dilutions ABL1 c.944C>T 10% 10.5% BRAF c.1799T>A 10% 13.1% CALR c.1092_1143del52 5% 7.0% CBL -1 c.1139T>C 10% 18.3% CBL -2 c.1259G>A 5% 17.7% VAF expected
FLT3-1 FLT3 -2 FLT3 -3 JAK2 -1 MYD88
Dup chr13:28,608,25028,608,277 (hg19) c.2503G>T c.1759_1800dup c.1624_1629delAAT GAA c.794T>C
VAF observed (50% Dilution) Mean 4.2% 5.7% 3.9% 6.7% 6.6%
VAF observed (25% Dilution) Mean neg neg neg 4.8% neg
VAF observed (12.5% Dilution) Mean neg neg neg neg neg
10%
8.5%
3.1%
1.1%
0.8%
10% 5%
13.6% 5.2%
6.1% 1.7%
neg 0.8%
neg 0.6%
10%
11.0%
6.7%
neg
neg
10%
9.8%
5.0%
neg
neg
neg
neg
neg
neg
neg
neg
neg
neg
Expected Variants detected in 1 or 2 of the 1 in 2 dilutions SRSF2
c.284_307del24 _
5%
7.4%
NPM1
c.863_864insTCTG
5%
4.5%
SF3B1 -1
c.2098A>G
5%
6.5%
SF3B1-2
c.1998G>T
5%
6.8%
No Expected Variants detected in any dilutions CSF3R c.1853C>T 5% IDH1 c.394C>T 5% JAK2 -2 c.1849G>T 5% MPL c.1544G>T 5% U2AF1 c.101C>T 10% CEBPA VAF 15% , low sensitivity CEBPA -2 c.939_940insAAG 15% Variants undetectable by Oncomine Myeloid ASXL1-2 c.1934_1935insG 10% CEBPA-1 c.68_69insC 15%
4.2%/neg/ 4.7% 4.1%/neg/ neg neg/4.0%/ neg neg/3.5%/ neg
6.8% 5.7% 7.7% 7.9% 10.0%
neg neg neg neg neg
neg neg neg neg neg
neg neg neg neg neg
6.7
neg
neg
neg
-
-
The LOD and the number of positive replicates differ between variant type.
Table 7 SeraSeq RNA control Accuracy.
Fusion Partners
Oncomine Myeloid Research Assay Primary Detected Isoform
Read count
MYST3-CREBBP ETV6-ABL1 (transcript 1) ETV6-ABL1 (transcript 2) PCM1-JAK2 FIP1L1-PDGRFA TCF3-PBX1 BCR-ABL1 RUNX1-RUNX1T1
KAT6A-CREBBP.K17C2 ETV6-ABL1.E4A2 ETV6-ABL1.E5A2 PCM1-JAK2.P23J12.COSF1001 FIP1L1-PDGFRA.F11P12del45 TCF3-PBX1.T16P3.COSF1489 BCR-ABL1.B14A2.1 RUNX1-RUNX1T1.R3R3
2895 3637 5657 5042 3903 2513 3364 4129
PML-RARA
PML-RARA.P6del11ins133A3
3942
Table 8 Performance characteristics on DNA clinical samples with different mutation types and difficult target genes.
Variant Type
Known Variants
Known Variants Detected
PPA (%)
PPV (%)
SNV indel (≤5bp) indel (5-40bp) indel (>40bp) FLT3-ITD* CEBPA* Total
50 21 15 6 6 12 92
50 18 15 6 6 11 89
100.0 85.7 100.0 100.0 100.0 91.7 96.7
98.0 94.7 100.0 100.0 100.0 100.0 97.8
Features are characteristic of samples with VAF>5%, Total=SNV + indel ≤5bp + indel (5-40bp) + indel (>40bp). *included in SNV/indel count; PPV=TP/(TP+FP)x100, PPA =TP/(TP+FN)x100, SNV, single nucleotide variants; indel, insertion/deletion, TP, true positive; FP, false positive; FN, false negative; PPA, Positive Percentage Agreement; PPV, positive predictive value, VAF-variant allelic frequency.
Table 9. Performance characteristics on RNA clinical samples with gene fusions.
Fusion Type
Primary Breakpoints detected PML(6) - RARA(3)
Secondary Breakpoints detected -
Cases
PPA(%) PPV(%)
2
100
100
1
100
100
11
100
100
100
100
100
100
1
100
100
1
100
100
t(15:17) PML(3) - RARA(3) t(8;21)
RUNX1(3) - RUNX1T1(3) BCR(14)-ABL1(2)
t(9;22)
BCR(13) - ABL1(2)
BCR(13) - ABL1(2) BCR(1) - ABL1(2) CBFB(5) - MYH11(33) CBFB(5) - MYH11(29)
inv(16)
-
CBFB(5) - MYH11(28) CBFB(4) - MYH11(29) CBFB(5) - MYH11(32)
CBFB(4) MYH11(33)/ CBFB(4) MYH11(29) CBFB(4) MYH11(28) CBFB(4) MYH11(32) -
KMT2A(9) - MLLT3(6) t(9;11) t(6;9)
DEK(9) - NUP214(18)
FIP1L1-PDGFRA
FIP1L1(13)_PDGFRA(12)
-
Total
6 2 9 2 1 1 1 1
KMT2A(9) MLLT3(6) -
KMT2A(10) - MLLT3(6)
2
2
43
Detected breakpoints are shown. Some samples had an additional secondary breakpoint.TPTrue positive; FN-False Negative, FP-False positive; PPA -positive agreement value; PPV – positive predictive value.
Table 10. DNA Panel Precision: three samples each with multiple variants were run in triplicates or duplicates in three runs.
Variant Allele Frequency Sample
Variant
Run 1
Run 2
Run3
Within Run Mean
0.45
0.42
0.42
Within Run SD
0.01
0.00
0.01
Within Run CV%
3.1%
0.0%
1.7%
JAK2 c.1849G>T (p.Val617Phe) 0.35 Within Run Mean
0.36
0.35
0.38
Within Run SD
0.04
0.01
0.01
11.8%
2.0%
1.9%
0.32
0.35
0.37
Between Run Mean
SD
CV%
0.43
0.02
4.4%
0.36
0.02
4.2%
0.35
0.03
7.7%
0.40
0.02
4.3%
0.55
0.01
1.8%
0.76
0.01
1.4%
0.18
0.01
5.0%
0.44
0.01
1.5%
SF3B1 c.2098A>G (p.Lys700Glu) 0.43
1
Within Run CV% NPM1 c.863_864insTCTG 0.37 Within Run Mean Within Run SD
0.12
0.01
0.01
38.0%
1.6%
3.1%
NPM1 c.870G>A (p.Trp290Ter) 0.38 Within Run Mean
0.40
0.39
0.42
Within Run SD
0.05
0.03
0.02
11.4%
6.5%
4.8%
Within Run Mean
0.56
0.54
0.54
Within Run SD
0.02
0.02
0.02
Within Run CV%
3.6%
2.8%
3.8%
Within Run Mean
0.77
0.75
0.76
Within Run SD
0.03
0.01
0.01
Within Run CV%
3.3%
0.8%
0.8%
CSDE1 c.35G>T (p.Gly12Val) 0.18 Within Run Mean
0.17
0.17
0.19
Within Run SD
0.01
0.01
0.02
Within Run CV%
5.9%
6.7%
11.2%
0.43
0.44
0.45
Within Run CV%
Within Run CV% CEBPA c.1009A>T (p.Thr337Ser) 0.56 2
STAG2 c.1777C>T (p.Gin593Ter) 0.73
NPM1 c.867_868insGAAA (p.Trp290fs) 0.44 Within Run Mean 3
Within Run SD
0.01
0.01
0.03
Within Run CV%
2.7%
2.3%
6.8%
IDH1 c.395G>A (p.Arg132His)
0.35 Within Run Mean
0.38
0.35
0.37
Within Run SD
0.03
0.02
0.02
Within Run CV%
8.1%
5.7%
4.7%
IDH2 c.419G>A (p.Arg140Gln) 0.11 Within Run Mean
0.09
0.10
0.09
Within Run SD
0.01
0.02
0.01
Within Run CV%
6.5%
16.6%
6.7%
CSDE1 c.34G>A (p.Gly212Ser) 0.069 Within Run Mean
0.07
0.07
0.07
Within Run SD
0.01
0.02
0.01
14.3%
21.4%
14.8%
Within Run CV%
0.37
0.01
3.8%
0.10
0.00
2.7%
0.07
0.00
2.7%
Within run variation inVAF is shown in rows, between run variation in columns.
Table 11. RNA Fusion Panel Precision: three samples each with gene fusions were run in triplicates or duplicates in three runs.
Sample
Fusion Detected
Sample1 Within Run Mean Within Run SD Within Run CV%
CBFB(5) - MYH11(29)
Sample2 Within Run Mean Within Run SD With Run CV% Sample 3 Within Run Mean Within Run SD Within Run CV%
Read Count
Between Run
Run1
Run2
Run3
Mean
SD
CV%
10573 470 4.45%
16093 3359 20.87%
7689 299 3.89%
11451
4271
37.3%
RUNX1(3)RUNX1T1(3) 33866 3244 9.58%
44000 7605 17.28%
28963 2251 7.77%
35609
7669
21.5%
17306 2681 15.5%
20210 2464 12.2%
16075 1101 6.8%
17863
2123
11.9%
PML(3) - RARA(3)
Within run variation in read counts shown in rows, between run variation in columns.
Table 12. RNA Fusion Panel sensitivity: Serial dilution of RNA Control down to 2 log reduction.
Fusion Partners
Read Count undiluted
Read Count 10-1
Read Count 10-2
MYST3-CREBBP ETV6-ABL1 (transcript 1) ETV6-ABL1 (transcript 2) PCM1-JAK2 FIPIL1-PDGRFA
2895 3637 5657 5042 3903
321 341 552 405 541
35 43 59 48 76
TCF3-PBX1 BCR-ABL1 RUNX1-RUNX1T1 PML-RARA
2513 3364 4129 3942
425 366 334 410
40 46 40 71