Integrating next-generation sequencing into clinical cancer diagnostics

Keywords::

• benchtop sequencing
• genomic medicine
• next-generation sequencing
• oncology diagnostics
• targeted sequencing

The advent of next-generation sequencing (NGS) has increased the throughput of DNA sequencing by >500,000-fold compared to traditional Sanger sequencing, while drastically plummeting the sequencing costs. The Human Genome Project completed a decade ago took 13 years and about US$3 billion to sequence the first human genome using shotgun method by Sanger sequencing. In contrast, by using NGS, sequencing a human diploid genome at 50–100 coverage costs about US$5000 nowadays and only takes a day or two [102]; ‘$1000 genome’ – a goal to be attained in the near future – will be a critical milestone in personalized genomic medicine where genome sequencing becomes an integral part of the clinical setting and will undoubtedly change the practice of medicine. In this article, we elaborate the developments of NGS technologies, and their applications in the oncology clinic. The limitations of NGS and the technical gaps as a diagnostic tool will also be discussed.

Evolution of NGS

The term ‘next-generation sequencing’ was coined to depict the evolutionary breakthrough in DNA sequencing technologies since the development of the chain-termination sequencing (Sanger sequencing) by Frederick Sanger in 1977. All NGS technologies are based on the model of massively parallel, high-throughput sequencing, which was first successfully demonstrated by the 454 sequencing platform (subsequently acquired by Roche) in 2005. Currently, the high-throughput NGS platforms are represented by Roche 454 GS FLX, Illumina GA/HiSeq and Life Technologies SOLiD and newly launched Ion Proton. It is noteworthy that although these are considered as high-throughput NGS platforms, the throughput of Roche 454 GS FLX (i.e., ~700Mb/run) is substantially lower than that of the other three platforms (i.e., >100Gb). The performances of the first three platforms have been extensively compared [1], while the performance of Ion Proton still needs to be evaluated. Despite their different sequencing chemistries and throughputs, all NGS platforms offer a sequencing cost at US$0.07–US$10/Mb in sharp contrast to US$2400/Mb by Sanger sequencing [1].

Of note are the availability of compact low/medium throughput sequencers in the past 2–3 years, exemplified by the Ion PGM, MiSeq and 454 GS Junior launched by Life Technologies, Illumina and Roche, respectively. Compared to the high-throughput NGS platforms, these benchtop sequencers offer a lower data throughput (from tens of megabases to several gigabases) at low costs, which are well suited for clinical applications. More importantly, in terms of the sequencing turn-around time, raw sequencing data are available within 3 h for the Ion PGM, while high-throughput NGS platforms typically take more than 1 week to complete a sequencing run. There is an increasing amount of data evaluating the clinical applications of these benchtop sequencers for detecting mutations in human cancer genes [2–4] with promising results.

Performance as a diagnostic tool

For diagnostic applications, the specificity and sensitivity of the benchtop sequencers are of critical concerns. In the clinical setting, false positive results may lead to erroneous decision for prophylactic surgery or drug administration, while false negative results may result in undiagnosed conditions. The benchtop sequencers differ in sequencing chemistries and mechanisms, hence the types of errors they generate. At the moment, MiSeq appears to be the most accurate among the benchtop platforms, whereas both Ion PGM and 454 GS Junior are susceptible to homopolymer errors, since they are based on a similar sequencing mechanism, resulting in an accuracy rate as low as 60% [5]. However, such homopolymer sequencing errors can be reduced by bioinformatic approaches such as error modeling [6] and selective filtering [7], without compromising the sensitivity. To date, a 100% sensitivity and 97% specificity have been reported for the Ion PGM [4]. Comparison of variant-calling accuracies between the platforms is not straightforward, as each manufacturer uses different software with unique quality scoring and detection threshold settings. Perhaps one way would be to standardize the sensitivity at 100% (or at an appropriate level such as 98%), and then compare the specificity of each platform. However, such comparisons are not widely reported in the literature.

Impact on cancer diagnostics

In the current clinical diagnostic realm, sequencing genes of interest for finding mutations, for example, in BRCA1/2 genes in hereditary breast and ovarian cancers, is still being performed by traditional Sanger sequencing, which can cost up to US$3700 and takes several weeks for the data to be generated. Impeded by the factors such as the lack of medical insurance coverage, the popularity of such analyses can be very low. In sharp contrast, switching from a Sanger-based to a NGS procedure will lead to about 60% reduction in reagent costs and turn-around time in a typical clinical setting [3]. Morgan et al. and Jiang et al. provided successful examples in developing PCR-based targeted NGS tests, using the Illumina GA and 454 Junior, respectively [8,9]. Morgan et al. sequenced TP53, BRCA1 and BRCA2 coding regions from 64 human DNA samples in one lane of the GA, reporting 100% sensitivity and specificity. Jiang et al. sequenced SEMA3A, SEMA3C and SEMA3D, genes implicated in Hirschsprung’s disease in 47 samples (in multiple runs), reporting a concordance rate of 97.7 and 95.6% in detecting known deletions and substitutions, respectively. Recently, Costa et al. validated the performance of Ion PGM for BRCA1 and BRCA2 >sequencing using a training set of 17 samples and a validation set of 20 samples, achieving a sensitivity of 100% and a specificity of 96.9% [4]. Together with plummeting sequencing costs, screening hundreds of genes within a timeframe relevant to tailoring treatment regimes is clearly a clinical and consumer reality.

Although NGS can be applied in detecting alterations at the genomic, transcriptomic (RNA-seq) and epigenomic levels, the most predominant application to date is in the sequencing of genomic DNA [10]. NGS provides a comprehensive view of different DNA aberrations, genetic recombination and other mutations unobtainable by any other technologies. One of the biggest potential applications of NGS is in the area of personalized genomic medicine, a medical model that proposes the customization of medical decisions and treatment regimes based on the individual patient’s genomic information. Personalized genomic medicine entails the use of companion molecular diagnostics – assays that measure expression levels of genes, specific mutations or other biomarkers so as to tailor medication to a patient’s specific needs. US Food and Drug Administration (FDA) has already made their position clear on companion diagnostics (in the July 2011 Draft Guidance) in which it expressed a desire for drug–diagnostic pairings to be approved together in the future. One of the recent examples of companion diagnostics is Qiagen’s ‘therascreen’ Kit used for detecting a KRAS mutation in codon 12 or 13, which needs to be absent in patient’s colorectal tumor in order to achieve therapeutic effect of the drug called cetuximab. NGS will undoubtedly offer exciting new opportunities for the development of companion diagnostics, as it has the capability of testing a large number of druggable gene targets simultaneously, for the purpose of selecting appropriate treatment for each patient. Not surprisingly, the number of clinical studies using NGS is on the rise [11].

There are multiple reasons for applying NGS in the oncology clinic. NGS can be used (a) for identifying germline mutations linked to familial cancer syndromes such as BRCA1/2 for familial breast and ovarian cancers and DNA mismatch repair genes for Lynch syndrome, or (b) for identifying somatic mutations in the genes which are of therapeutic importance within tumors such as EGFR for prescription of tyrosine kinase inhibitors. Although it is arguable that ultimately the cost for a full genome resequencing will become so affordable, making exome- and targeted-resequencing unnecessary, at the moment, the most logical step toward the widespread clinical implementation of NGS is the targeted approach to sequence a panel of genes, rather than whole genome or exome sequencing. Clinical implementation of NGS is likely to start as a simple replacement of existing Sanger sequencing or PCR-based assays, with the number of gene targets expanding progressively. For example, in May 2013, Myriad Genetics, the company that provides BRACAnalysis – the genetic test for germline BRCA mutations – announced the replacement of this test with a 25-gene NGS panel called myRisk by 2015 [103]. The myRisk panel will have a turnaround time of ≤14 days and a list price of about US$4,000 and will include genes such as BRCA1, BRCA2, BART, RAD51C, PALB2, PTEN, MYH and P16. It is also evident from the Foundation Medicine’s (Cambridge, MA, USA) NGS cancer panel, which screens for actionable mutations and rearrangements in 256 genes commonly altered in cancers [101]. Actionable alterations are defined as any alterations in a gene targeted by a therapy that is FDA approved for the cancer being tested or for other cancers, or currently in a clinical trial.

Concluding remarks

Although NGS has resulted in a massive explosion of genomic information relevant to cancer, challenges still remain for its translation into clinical practice in oncology. The technical challenges range from tumor heterogeneity in terms of cell types within tissue and genomic content within cells for somatic mutation detection; technical limitations to bioinformatics analysis and data or results interpretation. Furthermore, the ease of integration of NGS into current laboratory workflow will also depend on the types of samples employed for NGS. For example, for detection of somatic mutations, techniques employing FFPE samples are more easily integrated than those requiring fresh tissues. However, extraction of DNA from formalin-fixed paraffin-embedded (FFPE) samples can be very challenging.

Gullapalli et al. stated that bioinformatics is the single largest bottleneck in the clinical implementation of NGS [12]. Software has not reached a stage where it can be directly used in the clinic, as many NGS tools run on linux platforms (e.g., GATK, Samtools) requiring programming skills. Major computational concepts such as detection threshold and modeling of calling errors are still under experimentation. The scale of computation and demand for computer language literacy is still beyond the capability of typical clinical laboratories. New software and algorithms would have to be developed to efficiently and accurately analyze large NGS data sets, along with meaningful annotation and interpretation. A system of verifying and validating the reported results with an appropriate number of relevant samples would also have to be established. Finally, clinical laboratories deploying NGS assays would need a storage solution, in the form of either physical or cloud infrastructure, to transfer and store all NGS data for analysis, archiving and auditing.

The commercial success of NGS in diagnostic arena is largely subjected to other challenges beyond the technology itself, including regulatory approval, ethical concerns, the availability of insurance reimbursement and the cost of testing. The pace of technology development has outstripped the speed of the current regulatory process. The stance of FDA on the clinical applications of NGS is a crucial factor in shaping the future of NGS in clinical and diagnostic markets. Currently, there is no regulatory guidance or defined pathway for approval of NGS platforms and tests or the trial design needed for targeted resequencing of multiple genes. It would require extensive communication between FDA and all relevant parties to decide on the regulations. Once these technical and regulatory hurdles are overcome, there is little doubt that NGS will profoundly change the practice in medicine.

Acknowledgment

The authors would like to thank Zhen Xuan Yeo from the National Cancer Centre Singapore for his helpful discussions.

Financial & competing interests disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending or royalties.

No writing assistance was utilized in the production of this manuscript.