지난 10년 사이 차세대염기서열분석법은 급격한 속도로 발전하였고, 유전체학 연구 분야뿐만 아니라 임상의학 분야의 혁명을 일으켰다. 차세대염기서열분석법의 발전과 더불어, 분석적 타당도, 임상적 타당도, 임상적 유용성에 대한 근거들이 지속적으로 보고되고 있다. 현재 전 세계적으로 차세대염기서열분석법이 전면적으로 적용되고 있다. 본 종설에서는 진단검사의학과 의사와 임상병리사에게 필요한 임상유전체검사법의 도입을 위한 필수요소에 대한 리뷰를 제공하고자 한다. 먼저 현재 상용화된 차세대유전체검사장비의 장점 및 단점에 대해 간단히 리뷰하고, 임상유전체검사 검사과정 각 단계에서 발생할 수 있는 잠재적 오류 요인, 표준물질에 대해 논의하였다.
임상유전체분석법의 도입을 위한 필수 요소
Essential Elements for Establishing Clinical Next-generation Sequencing Testing
1Department of Laboratory Medicine and Biomedical Research Institute, Pusan National University Yangsan Hospital, Yangsan, Korea.
2Department of Laboratory Medicine, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Korea.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Lab Med Online 2019; 9(2): 37-44
Published April 1, 2019
Copyright © The Korean Society for Laboratory Medicine.
Over the past decade, next-generation sequencing (NGS) has evolved at an astonishing pace and has revolutionized clinical medicine as well as genomics research. The rapid advancements in NGS technologies have been accompanied by accumulating evidence of the analytical and clinical validity, and clinical utility of NGS. NGS is used worldwide. This review provides medical technicians and laboratory physicians with the essential elements for establishing clinical NGS testing. Here the authors briefly describe the advantages and drawbacks of currently available NGS platforms, potential sources of error in NGS workflow, and reference materials.
Over the past decade, next-generation sequencing (NGS), also known as massively parallel sequencing, has evolved at an astonishing pace and has revolutionized clinical medicine as well as genomics research [1, 2]. The now-global use of NGS has expanded the understanding regarding the mechanisms of pathogenesis of a variety of diseases, and has revealed a number of genetic alterations with diagnostic, prognostic, and therapeutic implications. Several examples include noninvasive prenatal test, newborn screening, Mendelian diseases, some heritable diseases, oncology, pharmacogenomics, infectious diseases, and metagenomics [3-6]. New clinical applications continue to be reported. The evidence to date has provided a compelling demonstration of the analytical validity, clinical validity, and clinical utility of NGS [5-8].
There have been several important milestones for clinical NGS implementation. The United States Food and Drug Administration (FDA) approved the MiSeqDx instrument along with two cystic fibrosis assays (Illumina, San Diego, CA, USA) and the Ion PGM Dx (ThermoFisher Scientific, Waltham, MA, USA) and Sentosa SQ301 (Vela Diagnostics, Fairfield, NJ, USA) in vitro diagnostic (IVD) tests (https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfRL/rl.cfm?lid=427645&lpcd=PFF and https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfRL/rl.cfm?lid=430009&lpcd=PFF). Several recommendations and guidelines to the implementation, validation, and reporting of NGS testing have been provided from professional organizations, including the American College of Medical Genetics and Genomics, Association for Molecular Pathology, Clinical and Laboratory Standardization Institute (CLSI), and federal agencies like the FDA and the US Centers for Disease Control and Prevention (CDC). Current Procedural Terminology medical code for NGS has been introduced into health care infrastructures worldwide. Proficiency testing (PT)/external quality assessment (EQA) programs for NGS have been implemented by accreditation bodies, such as the College of American Pathologists, European Molecular Genetics Quality Network, Korean Association of External Quality Assessment Service, and other professional organizations. The use of NGS by clinical laboratories worldwide continues to burgeon.
This review provides the essential elements for establishing clinical NGS testing, which will be helpful to medical technicians and laboratory physicians in the clinical laboratory setting. We discuss the advantages and drawbacks of currently available NGS platforms and focus on the potential error sources at each step throughout the clinical NGS workflow. Finally, quality control (QC) and reference materials (RMs) for NGS are described. QC aspects through PT/EQA programs have been addressed elsewhere [9, 10].
OVERVIEW OF NGS TECHNOLOGIES
It has been 13 years since NGS technologies appeared on the market. Compared to Sanger sequencing, which is considered the first-generation sequencing, NGS is characterized by multiple short reads generated from thousands-to-millions of sequencing reactions in parallel [1, 2]. Improvements in NGS technologies have in-creased the throughput and read length, and have reduced the cost and run time [1, 2]. Readers interested in the technological landscape and evolution of NGS platforms are directed to some previous reviews [1, 2]. This review briefly focuses on the differences between the platforms and platform-specific error profile (Table 1).
The current NGS platforms can be categorized according to three axes (Table 1). The first is cluster generation, which is the amplification vs. single molecule. The second is sequencing chemistry, which concerns the sequencing by synthesis (SBS, polymerase based) vs. sequencing by ligation (SBL) vs. nanopore. The third is detection method (optical detection vs. non-optical detection). When categorized by throughput, the NGS platforms can be categorized as high throughput systems, which include the HiSeq series and the NovaSeq series (Illumina), and benchtop systems, including MiSeq (Illumina), Ion Torrent Personal Genome Machine/Proton/S5 (ThermoFisher Scientific), and 454 GS Junior post-commercial platform (Roche, Branford, CT, USA). Each platform varies in terms of throughput, cost, read length, and run time.
Understanding the differences between these sequencing platforms is important, because some errors are specific to particular NGS platforms. For instance, the pH change in the Ion Torrent system or luminescence signal is imperfectly proportional to the number of nucleotides in case of homopolymer stretches (Table 1). The characteristics frequently result in homopolymer errors in semiconductor sequencing (e.g., Ion Torrent) and pyrosequencing. On the other hand, substitution error occurs more in Illumina platforms, because of sequence-specific interference .
While some errors are platform-specific, others, such as short read misalignment and amplification biases, are part of the shared error profile in most NGS methods. GC-rich or GC-poor regions are prone to be poorly amplified and are subject to false positive or false negative results and uneven coverage [11, 12]. In particular, for repeat or pseudogenes, short reads can be misaligned, which can reduce the number of on-target reads. In addition, short reads can lead to significant challenges in analyzing structural variations (SV) or copy number variations (CNVs) and complex RNA splicing patterns. Efforts to overcome these limitations prompted the de-velopment of the Illumina synthetic long-read sequencing platform (formerly known as Moleculo) and third-generation sequencing technologies, including those of Pacific Biosciences or the single molecule real-time sequencing technology (SMRT) of Oxford Nanopore [1, 2]. Although SMRT sequencing is less accurate than other platforms, the long reads that are obtained are advantageous for
NGS WORKFLOW, SOURCES OF ERRORS, AND QUALITY CONTROL
Clinical NGS implementations require quality management at multiple stages throughout the whole workflow. Most NGS testing workflows include sample preparation, library and template preparation, sequencing, data analysis, confirmation, clinical interpretation, and reporting (Fig. 1). Errors can occur at any level, and even low error rates in precedent stage could lead to fatal errors in the final results (Table 2). Early recognition and detection of errors is essential to avoid risk for delayed turnaround time and, more importantly, to exclude false positive and false negative results and misdiagnosis. Expected sources of errors, QC metrics, and optimization in NGS testing are summarized in Table 2.
Figure 1. Overall workflow of next-generation sequencing.
1. Sample preparation
The performance of the NGS testing depends on the purity and concentration of the nucleic acids, which widely vary according to specimen types, which include formalin-fixed paraffin-embedded (FFPE) tissues, fresh tissues, blood, bone marrow, and cell-free DNA. At this stage, nucleic acid degradation, presence of inhibitors, contamination (e.g., microorganisms), and carry-over could be important sources of error [13, 14]. It is also important to avoid clerical errors from mislabeling. Furthermore, in case of tumor samples, tumor burden, sequencing of matched-normal samples, and formalin preparation should be considered . The tumor cell fraction could influence detection sensitivity of, in particular, CNVs and estimation of rare mutant allele frequencies (AF). Tumor-only sequencing is prone to lead to numerous false positive and false negative results . FFPE tissues are problematic due to artifactual sequence changes (formaldehyde-induced crosslinks, extensive DNA fragmentation, abasic sites, and deamination of cytosine bases) [13, 14]. Therefore, acceptance and rejection criteria, which include the absorbance ratio at 260/280 nm, DNA/RNA integrity number, and tumor burden, for specimens should be determined in clinical laboratories (Table 2).
2. Library and template preparation
Library construction for current NGS involves fragmentation, adapter ligation, enrichment, and clonal generation. DNA fragmentation can be obtained by mechanical methods, such as nebulization and sonication, or by enzymatic digestion. After the end repair of fragmented DNA, one or more platform-specific adapters are ligated to the fragmented DNA. Adapters provide sequences for binding to primers as well as index sequences for multiplexing. Enrichment is then performed by target amplification or target capture by hybridization in targeted resequencing (Table 3). Lastly, DNA templates are clonally amplified on beads (e.g., emulsion PCR, ePCR) and on a solid surface (e.g., bridge amplification) in parallel, and millions of clonal clusters are created [1, 2].
There are several important sources of errors to occur during library construction. First, DNA fragmentation is not actually completely random and varies according to the GC content or genomic location. This could promote preferential amplification and introduce coverage biases. Second, adapter dimers and index swap result in decreased data volume and contamination of datasets, respectively. When there is a lack of barcode diversity for sample multiplexing, false positive calls can be incorrectly assigned to different samples due to index swap. The swap rate could be increased when shorter index sequences are used and index mismatches are allowed during demultiplexing. Third, primer biases and PCR-mediated errors such as PCR stochasticity, polymerase errors, and PCR template switches could occur [14, 16]. Sequencing results would be confounded by allele dropout in case with variants in primer binding sites. Potential error sources (disadvantages) of target amplification and capture by hybridization are described in Table 3. Lastly, overclustering in bridge ampli cation leads to loss of focus in image analyses and decreases in data quality . In contrast, underclustering results in lower output. Similarly, an inappropriate bead-to-fragment ratio in ePCR might lead to multi-clonal generation.
To identify the error source at this level, distribution of library size should be checked to determine whether there are fragments with the expected size. Sometimes, there may be abnormal peaks due to adapter dimer. Accurate quantification of the library is critical for optimal cluster density and monoclonal amplification (Table 2).
3. Sequence generation
The sequences are determined by SBS, SBL, and nanopore. Considering the current market share of the platforms, this review focuses on SBS methodology. SBS consists of cyclic reversible termination (CRT), single nucleotide addition (SNA), and real-time sequencing. In CRT approaches, a mixture of all four labeled nucleotides is added. Sequencing involves the incorporation of each fluorescently labeled reversible terminator nucleotide per cycle (e.g. Illumina system and Qiagen GeneReader system) [1, 2]. On the other hand, in SNA, each of the four nucleotides is added, and sequencing involve cyclic incorporation of unlabeled nucleotides (ThermoFisher Scientific Ion Torrent/Proton/S5 system) [1, 2]. Unlike CRT, SNA does not require terminator nucleotide, because the absence of the nucleotides prevents DNA strand elongation.
One source of error at this level involves nucleotide incorporation by polymerase. Uniformity of the sequencing signal, number of bases sequenced, and sequencing accuracy can be determined by the performance of polymerase. Another error can arise from artifacts of the fluorophores used as the detection signal. Poor quality fluorophore is related to background noises and overlapping signals, which leads to incorrect base calls. As stated above, inherent technical limitations include interferences due to homologous sequences, high or low GC contents, and homopolymer errors.
Considering the size of the genomic data, even the lowest error rate could give rise to thousands of false positive variants. The accuracy or error rate of the sequence data is presented as the Q-score, which is a logarithmic value of the probability of an incorrect base call. The Q-score differs depending on the sequence context and the platforms. For instance, Q-scores tend to decrease near homopolymer stretches or indels in reads. Reported error rates for the NGS platforms differ considerably, because different NGS platforms calculate Q-scores using different algorithms (Table 1). The Q-scores from sequencer-derived raw data are often nonlinear and subject to overestimation. Thus, the Q-score should be recalculated by the process of base quality recalibration considering the position of the base in a read (sequencing cycle), the preceding sequence, and the error rate of the sequencer. Before data analysis, it is important to review quality metrics of raw data including Q-scores, per base GC contents, per base N content, and sequence duplication levels as well as basic statistics (Table 2).
4. Data analysis
NGS data analyses requires substantial computational infrastructure due to the large amounts of data generated and short reads. Although the bioinformatics pipeline varies according to the application, it generally consists of three steps (Fig. 1): primary, secondary, and tertiary analysis. Primary analysis begins with the generation of base calls from fluorescent image or electrical current signals and generates a FASTQ file. Secondary analysis is the process of mapping to a reference sequence or de novo assembly and variant calling. The file formats of read alignment and variant calling are sequence alignment/Map (SAM)/Binary alignment/MAP (BAM) and variant call format, respectively. Finally, tertiary analysis involves the variant annotation and prioritization. A variety of bioinformatics tools have been reported, and the number of tools is still expanding [18-20]. Unfortunately, anal-yses results using different tools still have poor concordance . Readers interested in the collection of bioinformatics tools are referred to previous reviews [18-20].
Sequencing errors can be also observed at any level of the bioinformatics analyses. Quality trimming is the starting point of bioinformatics analyses that aims to improve the accuracy of the results. Although quality trimming can lead to marked decrease in coverage, in particular exome data, this process could improve the mappability of short reads. As stated above, short reads can result in misalignment and inaccurate variant calls. Choice of mapping software and completeness of reference genome can influence the accuracy of the read mapping. In addition, for a read containing multiple variants, a read that spans a deletion or duplication, or for a sequence that is not present in the reference genome, a read cannot map to reference genome . Conversely, there is a situation where reads map to multiple locations because reads are derived from highly homologous regions [21, 22]. Thus, homologous sequences including pseudogenes may result in false positive or false negative results. Erroneous calls can often arise from the alignment of indel-containing reads, because an indel can increase the number of mismatches [13, 21]. This error can be corrected by an additional process, such as ‘local realignment’ around a set of known indels [13, 19].
In addition to mapping quality, erroneous calls can arise from strand bias (imbalance of the number of forward and reverse reads), excessive duplicate reads, or low depth of coverage. Duplicate reads, which generally result from amplification, can bias the determination of AF. Moreover, it is difficult to distinguish true variants with low AF and erroneous calls, while low level variants may not be detected under a low depth of coverage [22, 23]. Thus, NGS data analyses may create errors depending on the underlying sequence context and experimental design, as well as the bioinformatics tools. However, considering the inherent differences among NGS platforms, wet experiments, and bioinformatics pipelines, the threshold or cutoff of quality metrics cannot be specifically defined (Table 2).
REFERENCE MATERIALS FOR NGS
Considering the potential error sources in NGS testing, the use of reference materials (RMs) is essential to implement and validate NGS testing. Some errors can be mitigated by increasing the depth of coverage, supplementing with dual indexed barcoding, and using replicates [14, 23, 24]. However, systematic errors due to sequencing artifacts or misalignment cannot be corrected without RM [14, 23]. In addition, well-characterized RM can be useful for performance evaluation, validation of analytic pipeline, internal QC, PT/ETQ program, and calibration of quantitative measurements, such as AF, CNV, or expression abundance.
As stated above, NGS testing consists of a multifaceted workflow and produces a huge number of variants and potential errors in different genomic contexts. Unlike RM in other genetic testing, one pitfall in RM for NGS is the requirement for materials that contain the full spectrum of variants (SNV, indels CNV, and SV) and a variety of variant AF in different genomic contexts. To address this issue, the Genome in a Bottle Consortium was hosted by National Institute of Standards and Technology (NIST) in 2012. Subsequently, in 2015, NIST released the first human genome RM 8398, which is genomic DNA (NA12878) derived from the Coriell cell line GM12878. Reference values for RM8398 consist of single nucleotide polymorphisms, small indels, and homozygous reference genotypes for approximately 77% of the genome. In 2016, three additional human RMs, including RM 8391 (GM24385, male of Eastern European Ashkenazic Jewish), RM 8392 (GM24149, GM24143, trio of Eastern European Ashkenazic Jewish), and RM 8393 (GM24631, male of East Asian) became available (http://jimb.stanford.edu/giab/).
RMs for NGS can be classified into three types . The first is a well-characterized cell line, such as NIST RM. The second is a synthetic RM, such as plasmid spike in the genome and a single mix of multiple synthetic oligo. The third is
Although advances in NGS technology have revolutionized medicine, significant errors can still occur. The review has outlined advantages and drawbacks of NGS platforms, the range of error sources throughout the NGS workflow, and RMs for NGS. The issues described here, combined with other validation guidelines from professional societies, can be helpful for medical technicians and laboratory physicians who implement and validate NGS testing in clinical laboratories.
AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
No potential con icts of interest relevant to this article were reported.
- Levy SE, Myers RM. Advancements in next-generation sequencing. Annu Rev Genomics Hum Genet 2016;17:95-115.
- Goodwin S, McPherson JD, McCombie WR. Coming of age: ten years of next-generation sequencing technologies. Nat Rev Genet 2016;17:333-51.
- Minear MA, Alessi S, Allyse M, Michie M, Chandrasekharan S. Noninvasive prenatal genetic testing: current and emerging ethical, legal, and social issues. Annu Rev Genomics Hum Genet 2015;16:369-98.
- Park KJ, Park S, Lee E, Park JH, Park JH, Park HD, et al. A population-based genomic study of inherited metabolic diseases detected through newborn screening. Ann Lab Med 2016;36:561-72.
- Bodian DL, Klein E, Iyer RK, Wong WS, Kothiyal P, Stauffer D, et al. Utility of whole-genome sequencing for detection of newborn screening disorders in a population cohort of 1,696 neonates. Genet Med 2016;18:221-30.
- Pritchard CC, Salipante SJ, Koehler K, Smith C, Scroggins S, Wood B, et al. Validation and implementation of targeted capture and sequencing for the detection of actionable mutation, copy number variation, and gene rearrangement in clinical cancer specimens. J Mol Diagn 2014;16:56-67.
- Lih CJ, Sims DJ, Harrington RD, Polley EC, Zhao Y, Mehaffey MG, et al. Analytical validation and application of a targeted next-generation sequencing mutation-detection assay for use in treatment assignment in the NCI-MPACT trial. J Mol Diagn 2016;18:51-67.
- Saunders CJ, Miller NA, Soden SE, Dinwiddie DL, Noll A, Alnadi NA, et al. Rapid whole-genome sequencing for genetic disease diagnosis in neonatal intensive care units. Sci Transl Med 2012;4:154ra135.
- Segal JP. Next-generation proficiency testing. J Mol Diagn 2016;18:469-70.
- Aziz N, Zhao Q, Bry L, Driscoll DK, Funke B, Gibson JS, et al. College of American Pathologists' laboratory standards for next-generation sequencing clinical tests. Arch Pathol Lab Med 2015;139:481-93.
- Aird D, Ross MG, Chen WS, Danielsson M, Fennell T, Russ C, et al. Analyzing and minimizing PCR amplification bias in Illumina sequencing libraries. Genome Biol 2011;12:R18.
- Benjamini Y, Speed TP. Summarizing and correcting the GC content bias in high-throughput sequencing. Nucleic Acids Res 2012;40:e72.
- Jennings LJ, Arcila ME, Corless C, Kamel-Reid S, Lubin IM, Pfeifer J, et al. Guidelines for validation of next-generation sequencing-based oncology panels: a joint consensus recommendation of the Association for Molecular Pathology and College of American Pathologists. J Mol Diagn 2017;19:341-65.
- Robasky K, Lewis NE, Church GM. The role of replicates for error mitigation in next-generation sequencing. Nat Rev Genet 2014;15:56-62.
- Jones S, Anagnostou V, Lytle K, Parpart-Li S, Nesselbush M, Riley DR, et al. Personalized genomic analyses for cancer mutation discovery and interpretation. Sci Transl Med 2015;7:283ra53.
- Kebschull JM, Zador AM. Sources of PCR-induced distortions in high-throughput sequencing data sets. Nucleic Acids Res 2015;43:e143.
- Illumina. https://support.illumina.com/content/dam/illumina-marketing/documents/products/other/miseq-overclustering-primer-770-2014-038.pdf (updated on August 2017).
- Pabinger S, Dander A, Fischer M, Snajder R, Sperk M, Efremova M, et al. A survey of tools for variant analysis of next-generation genome sequencing data. Brief Bioinform 2014;15:256-78.
- Oliver GR, Hart SN, Klee EW. Bioinformatics for clinical next generation sequencing. Clin Chem 2015;61:124-35.
- Chiara M, Pavesi G. Evaluation of quality assessment protocols for high throughput genome resequencing data. Front Genet 2017;8:94.
- Gargis AS, Kalman L, Bick DP, da Silva C, Dimmock DP, Funke BH, et al. Good laboratory practice for clinical next-generation sequencing informatics pipelines. Nat Biotechnol 2015;33:689-93.
- Santani A, Murrell J, Funke B, Yu Z, Hegde M, Mao R, et al. Development and validation of targeted next-generation sequencing panels for detection of germline variants in inherited diseases. Arch Pathol Lab Med 2017;141:787-97.
- Hardwick SA, Deveson IW, Mercer TR. Reference standards for next-generation sequencing. Nat Rev Genet 2017;18:473-84.
- Schmitt MW, Kennedy SR, Salk JJ, Fox EJ, Hiatt JB, Loeb LA. Detection of ultra-rare mutations by next-generation sequencing. Proc Natl Acad Sci U S A 2012;109:14508-13.
- Sims DJ, Harrington RD, Polley EC, Forbes TD, Mehaffey MG, McGregor PM 3rd, et al. Plasmid-based materials as multiplex quality controls and calibrators for clinical next-generation sequencing assays. J Mol Diagn 2016;18:336-49.
- Kudalkar EM, Almontashiri NA, Huang C, Anekella B, Bowser M, Hynes E, et al. Multiplexed reference materials as controls for diagnostic next-generation sequencing: a pilot investigating applications for hypertro-phic cardiomyopathy. J Mol Diagn 2016;18:882-9.
- Illumina. https://www.illumina.com/content/dam/illumina-marketing/documents/products/technotes/hiseq-phix-control-v3-technical-note.pdf (updated on August 2017).
- Duncavage EJ, Abel HJ, Merker JD, Bodner JB, Zhao Q, Voelkerding KV, et al. A model study of in silico proficiency testing for clinical next-generation sequencing. Arch Pathol Lab Med 2016;140:1085-91.
- Davies KD, Farooqi MS, Gruidl M, Hill CE, Woolworth-Hirschhorn J, Jones H, et al. Multi-institutional FASTQ file exchange as a means of proficiency testing for next-generation sequencing bioinformatics and variant interpretation. J Mol Diagn 2016;18:572-9.