Editor’s Note: In this 2-part guest blog series, Keith Robison, PhD (Lead Senior Scientist, Informatics, Infinity Pharmaceuticals) gives a reacap of the session he chaired at last month's Applying Next-Gen Sequencing conference. Keith also maintains the popular Omics! Omics! blog.
October 20, 2011
Keith Robison : (Part 2 of 2) The following joint session was on next-next-generation sequencing, the possible technologies of the future. Two of the talks were intriguing but describing very distant futures: single-molecule reversible terminator sequencing with optical tweezers and an optical variant on nanopore sequencing. The talk I would have love to have yanked into my session was a natural follow-on, with John Healy of GnuBIO describing their progress on a platform they hope to launch in the first half of next year.
In my talk I described a process of PCR amplifying samples in a thermocycler and then shipping them off to a vendor to run on the PGM. Barcoding was a natural inclination, given a sometime need to analyze multiple samples and amortize the run cost over those many samples. Gnu’s vision compresses this and simplifies it, as a single instrument both performs the amplifications and analyzes them, returning to the user a list of differences rather than data which must be processed further. All from a box slated to cost $50K and with consumables in the $50-$100 per sample range with sample-to-data times projected at 3 hours.
Gnu’s approach is based on picodroplet technology and continues the trend of pushing most of the guts of the technology into the disposable cartridge. Microfluidic channels in the cartridge convert samples to large droplets which are merged with droplets bearing primer pairs; imagine one of those droplet-generating desk toys shrunk a few zillion-fold. A serpentine channel moves each such amplification droplet through hot and cold zones to perform PCR. Each droplet is then split further into smaller droplets, which are in turn merged with detection droplets. Each detection droplet contains a specific hexameric detection probe, and is also color-coded. After performing a primer extension reaction, the color codes and signal are read. All of the microfluidics are in the disposable card, with new applications requiring new card designs.
The detection chemistry is a simple variant on sequencing-by-hybridization (SBH). Hybridized hexameric probes are extended, resulting in the activation of a dye molecule. GNU is aiming this towards the resequencing market, which greatly simplifies SBH. By examining both the positive and negative signals and comparing against the known sequences of the amplicons, variants can be detected as unusual signatures. This is very much easier than attempting de novo SBH, which must reconstruct the sequence from the signals. Furthermore, the system should be robust to a certain degree of false positive and false negative probes; the complete ensemble of probe signals can be used to weight various models of what sequence the amplicon contains.
Note also that the signals are purely binary; presence of two copies of a hexamer does not lead to a different signal from an amplicon containing only one. The “read length” of the system is currently approaching a kilobase, but it is important to note the differences from reads on other platforms. SBH cannot read any repeat longer than the probe length (6), and it is easy to construct variant scenarios that cannot be distinguished. Read lengths here are really amplicon lengths, with the limitation being the complexity of the probe library required. Healy proposed that another four-fold increase in amplicon size by lengthening the probes to heptamers and using four cards, each with a different quarter of the library, to sequence these. But, while these are important caveats, they simply rule out a small number of genotyping or mutation detection applications. Also, in the cancer space one generally works with FFPE DNA, which is inherently fragmented to about 450 base pairs; GNU is already reading much longer than this.
A number of strengths from this approach are apparent. In my PGM experiments differential primer performance led to different sensitivities for different amplicons; in GNU’s approach each amplicon is analyzed in its own picodroplet, eliminating cross-talk. The proposed cost of the consumables pretty much eliminates the temptation to bar-code; why bother when the run cost is so small? However, on the negative side the current system has a minor allele frequency somewhere around 20-30%, about that of Sanger sequencing. This is because each original amplification droplet is expected to carry many copies of each amplification template. However, in response to my question on this topic, Healy commented that by diluting the sample so each droplet contains about half a genome, then very high sensitivity should be achievable. It is also easy to imagine how the system could be used for other detection modalities, such as single-base extension or a probe ligation assay or just a probe-hybridization assay. The system would also naturally be able to slide into digital PCR. Could COLD-PCR be adapted to this platform? That would require the hinted common thermal profile across amplicons, but once this is accomplished it would appear to be a natural fit and another possible approach to detecting rare alleles.
Whether any or all of this will come to pass will depend on successful launch of their platform next year. Until then and probably continuing afterwards, many will be performing experiments similar to mine to develop the next generation of cancer mutation assays, as well as using rapid platforms such as PGM and MiSeq to follow-up on discoveries from germline and somatic exome or whole-genome sequencing.
Read Part 1