October 1, 2016 (Vol. 36, No. 17)
The next-generation sequencing (NGS) market has been and continues to be dominated by Illumina, causing many researchers to complain that the lack of competition is causing the rate of NGS advancement to slow, at least in terms of cost for short-read sequencing.
Since the launch of Illumina’s flagship HiSeq X sequencing platform, the cost for a research-grade whole genome has held steady at around $1,000.
The only potential competition on the horizon, BGI’s Revolocity (based on technology from Complete Genomics), has since been shelved. However, there have been a number of exciting developments that promise to open up the use of NGS in other areas.
One area in which tremendous progress is being made is that of long reads, which are revealing areas of complexity in the genome and transcriptome that defy short-read sequencing. In the analysis of genomes, the primary problem is dealing with repeated sequences of DNA. Repeats longer than the underlying read length can be very difficult or even impossible to uniquely map to the reference genome.
Similar challenges occur in transcriptome analysis. “Most genes contain multiple alternative exons that are located farther apart in the mRNA than the read length,” says Brenton Gravely, Ph.D., the John and Donna Krenicki Professor of Genomics and Personalized Healthcare at UConn Health. “It is, therefore, impossible to use it to accurately know which isoforms are present in a particular sample.”
The company that’s been getting the most attention lately around long reads is Oxford Nanopore, which first made their nanopore-based sequencer, the MinION, available last spring. Although Oxford Nanopore has stretched the definition of a formal product launch, the company has certainly made its technology available to a broad audience. Over 1,000 research groups are using the MinION in their own laboratories.
Within a year following its initial release, the MinION increased its output from less than 100 Mb to over one Gb, primarily through a series of improvements in nanopore design and enzyme chemistry. And while the MinION hardly approaches what Illumina’s platform offers in terms of output, quality, and cost per Gb, the Oxford Nanopore instrument has two main attractions: read length and portability.
Users have reported individual reads of over 100 Kb, with average read lengths in the many tens of thousands of bases. DNA input quality and library prep appear to be the only real factors limiting the read length—if you feed the nanopores long DNA molecules, you will get long reads.
Because the MinION is a handheld sequencer (with fans arguing over the best size comparator—stapler, USB stick, smart phone, chocolate bar, etc.), it is even possible to bring the sequencer out to the field where the data is being collected, rather than processing samples after they have been brought back to the laboratory.
Data quality lags far behind Illumina, but it is improving over time. The latest R9 pore in “fast mode” is getting accuracy rates up to 95% under the best conditions.
Pacific Biosciences, the leader in long-read sequencing, has been generating some excitement of its own. The company announced a major new platform, the Sequel, at ASHG last year. This platform, which is the result of a collaboration with Roche, promises several improvements over the machine it is replacing, the RS II.
The Sequel, in comparison with the RS II, generates sevenfold greater output, occupies two-thirds less space, and requires one-half the capital costs. Pacific Biosciences has struggled a bit to get the platform fully launched, dealing both with chemistry issues and supply issues, especially for the consumable SMRT (single molecule, real time) cells. Both issues appear to have been resolved, so we are anxiously awaiting the first customer-generated data.
The other player to make waves in this space is 10X Genomics, with its GemCode™ technology, which can generate linked reads from underlying short-read data. Although this technology doesn’t have all the benefits of true long reads (for example, it cannot help with repeats that are longer than the underlying short reads), the linked reads are able to generate substantially better de novo assembly and phased reads.
The platform appears to work well enough that Illumina signed a co-marketing agreement with 10X Genomics earlier this year, effectively if quietly signaling the death of Illumina’s own Moleculo synthetic long-read technology.
Despite all of the progress, there are still issues with long-read technology, especially for transcriptome sequencing. “One of the issues is the current throughput of these platforms, but the bigger issue is that reverse transcriptases are just not processive enough,” explains Dr. Gravely. “A really great reverse transcriptase—or even better—direct RNA sequencing, would be game changers.” Transcriptome researchers will be happy to note that Oxford Nanopore, which recently published a paper on using the MinION to directly sequence RNA, has plans to launch a commercial kit in the future.
Another application that has been gaining a lot of traction is single-cell genomics. Until recently, all NGS projects (with only rare exceptions) were performed on a mixture of cells, anywhere from thousands of cells up to one million cells for a single sequencing library. If all of the cells are completely homogeneous, as is generally the case for somatic cell genomes, this isn’t really a problem. However, there are several applications where the cells are not homogeneous, and pooling them together masks critical information.
For example, tumor biopsies are notoriously heterogeneous, comprising both somatic and cancer cells. And even the cancer cells from the same tumor can have substantially different genomes. Pooling multiple cells together will produce a mixed genomic profile, complicating the interpretation. Variations present in only a small number of cells may be completely obscured.
A strategy some are starting to attempt is to take several single-cell measurements from different locations within the biopsy. If enough individual measurements are made, a comprehensive view of the tumor genome can be built up.
Transcriptomics is another application that is starting to benefit from the single-cell approach. The transcriptome can be highly variable from cell to cell, and pooling cells together can obscure the underlying variability in gene-expression patterns.
“Only a few years ago, the transcriptomics analysis of a few hundred single cells would have taken significant resources and time,” notes Richard Shen, Ph.D., a genomics scientist and entrepreneur who recently left Illumina to found RS Technology Ventures. “Now, with the development of easy-to-use NGS library preparation methods and decreases in sequencing costs, the analysis of a few thousand single cells is not only possible but also desirable in many applications.”
For the preparation of single-cell libraries, several platforms are available, including the C1 from Fluidigm, which can process 800 cells at a time, and the Chromium™ from 10X Genomics, which can process up to an astonishing 48,000 cells at a time. Such processing capabilities, however, may not be to every investigator’s purpose, as Dr. Gravely notes: “The throughputs of some of the new platforms or homemade devices are very exciting, but for someone who works on splicing, the holy grail is single-cell, long-read sequencing.”
Cancer Liquid Biopsies
Yet another notable NGS development is the introduction of liquid biopsies for cancer diagnosis. The use of NGS is becoming somewhat routine in the diagnosis of cancer tumor biopsies, with Foundation Medicine’s FoundationOne test being a prime example—a targeted panel of 315 cancer-related genes combined into a single test. These panel-based tests are replacing the more traditional, single-gene tests.
Cancer diagnostics are also shifting from tumor biopsies to liquid biopsies. Instead of extracting DNA from a section of the tumor, cancer liquid biopsies involve extracting either circulating tumor cells (CTCs) or cell free DNA (cfDNA) from the patient’s blood. While solid tumor biopsies won’t go away due to the anatomical and cellular structure information they provide, there’s a lot of excitement around liquid biopsies because of the versatility they offer.
“From a cancer-screening perspective, liquid biopsies are attractive because they represent a simple and accurate (potentially) presymptomatic test that could be performed routinely to catch cancer early,” states Gabriel Otte, CEO and co-founder of Freenome, a liquid biopsy company. “From a prognostic or patient-monitoring perspective, they are interesting for similar reasons, but especially for up to 30% of cases of cancer that cannot be invasively biopsied for a variety of reasons.”
As might be expected in a hot new field, several new companies are being formed, and more established companies are retooling to take advantage of what is expected to be a large and growing market. At AllSeq we are tracking over 35 companies in this space.
Guardant Health claims that its cancer liquid biopsy diagnostic was the first to reach the market when it was launched in early 2014. This diagnostic, called Guardant360, looks at 70 cancer genes. Freenome is taking a different strategy by sequencing the entire cancer genome. “We rely on our deep learning,” says Otte, “to figure out the specific subregions that are most relevant for making an accurate cancer test.”
Illumina, whose sequencing technology is being used by most of these companies, has decided it wants to get in the game as well by spinning out its own liquid biopsy company, Grail. However, perhaps in an effort to avoid the appearance of competing directly with its own customers, Illumina reports that Grail will be focusing on presymptomatic screening, a much more challenging task that none of the other companies has yet to focus on.
Looking to the Future
While Illumina looks to continue dominating the market for high-throughput short reads, we are keeping our eyes out for advancements in other areas. Oxford Nanopore has announced a lot of upcoming launches, including its high-throughput nanopore sequencer, the PromethION. The company claims that it will have the throughput to compete with Illumina’s HiSeq X.
The PromethION is currently in the hands of at least a couple of early-access customers, but there is no firm launch date yet. Oxford Nanopore is also talking about two instruments for automating sample handling, the VolTRAX and the Zumbador, which promise to further simplify the whole process of sequencing DNA samples.
We will also be looking to see how quickly Pacific Bioscience’s Sequel platform gets adopted. It may start bringing the cost of long reads down enough to pull some business away from Illumina. Finally, we’re keeping an eye on clinical applications of NGS. With steady improvements in ease of use and reductions in cost, the barriers to clinical adoption now tend to center more on regulation and reimbursement.
Improving Sample Tracking in Longitudinal Studies
With the advent of liquid biopsy assays to monitor treatment response of oncology patients in research studies, proper tracking of samples has become increasingly critical. At the onset of such programs, a minimum of three samples are analyzed for their genetic profile; tumor tissue, normal tissue, and circulating, cell-free DNA.
When performing sequencing on a large target such as whole exome sequencing, a genetic fingerprint can be determined to confirm that samples are properly matched during data analysis. However, when sequencing with small targeted panels, a genetic fingerprint cannot be produced, according to Drew McUsic, Ph.D., an application scientist at Swift Biosciences.
“Until now, researchers have utilized single nucleotide polymorphism based arrays paired with LIMS systems to properly track samples,” noted Dr. McUsic. “These approaches rely on proper labeling of multiple samples and data files to maintain exact matches of all materials.”
Searching for an integral, more precise method of sample identification, Swift Biosciences developed the Accel-Amplicon™ Sample_ID panel. A genetic fingerprint provided by the 104 exonic and gender specific amplicons is ideally utilized as a low percentage spike-in to any Swift amplicon-based panel, such as the Accel-Amplicon 56G Oncology Panel v2, explained Dr. McUsic, adding that this results in sample identification from low depth sequencing of germline targets while still enabling high depth of coverage for somatic mutation detection.
“Such a technique provides an efficient, single-tube assay for analyzing somatic mutations in oncology specimens while generating the genetic fingerprint within the same sequence file,” he continued. “As more projects are designed to track response rates in oncology patients, it will be critical to select proper sample tracking tools in these longitudinal studies.”