This article was originally published February 14 and has been modified to include additional comments.
MARCO ISLAND, Fla. — The Broad Institute's David Jaffe presented the first data from Oxford Nanopore Technologies' MinIon nanopore sequencer last week at the Advances for Genome Biology and Technology meeting in Marco Island, Fla.
Jaffe discussed using the reads from the nanopore sequencer to disambiguate assemblies created with Illumina sequence data.
Oxford Nanopore generated MinIon data from two bacterial genomes — a 4.7 megabase Escherichia coli genome and a 1.6 megabase Scardovia genome for which the Broad has a finished reference genome.
The MinIon chip consists of a 1 centimeter squared rectangular array with 512 nanopores. Jaffe said that the speed at which DNA passes through the pore is configurable and ranges between one base per second to 100 bases per second. Oxford sequenced the two bacterial genomes at a speed of 25 bases per second.
Oxford sequenced the E. coli genome to six-fold coverage, obtaining a mean read length of 5.4 kilobases and sequenced the Scardovia genome to 13-fold coverage with a mean read length of 4.9 kilobases.
Previously, Oxford said that its systems would be able to deliver read lengths of up to 100 kilobases.
Jaffe said that when he asked Oxford why the read lengths were so short, Oxford said that the reads were limited by the sample and the initial fragmentation.
In regards to accuracy, Jaffe said that 84 percent of the reads at 5 kilobases or longer had at least one perfect 50-mer, while 100 percent of the 5-kb or longer reads had at least one perfect 25-mer.
Some of the errors seem to be a function of the fact that several bases are present in the nanopore at one time.
According to Jaffe, the MinIon nanopore system consists of a protein nanopore with a ratcheting enzyme attached to it. Double-stranded DNA is tethered to the membrane on one end and bound to the ratcheting enzyme on the other end. The ratcheting enzyme strips bases from the double-stranded molecule, feeding a single strand of DNA through the nanopore.
As the DNA passes through the pore, current flowing across the membrane generates a signal. Because there are several bases in the pore at any given moment, the current that is generated is the function of several bases, or k-mer, which he said are 6-mers. That current signal is then translated into the equivalent k-mers, which are then translated into reads.
Jaffe said that looking at the sequence data, there were "long perfect stretches," but also "blocks of errors every once in a while." These error blocks appear to occur when translating the current signal into a k-mer. "Something goes wacky and it's out of sync for a while," he said. The system then rights itself and begins reading correctly again.
Additionally, he said, there appears to be artifacts that are due to the nanopore itself. For example, when looking at multiple alignments of six E. coli reads at a base location where the true answer is T, the MinIon called it as a T in only one of the six reads.
Jaffe said that these types of errors could be decreased by improved base calling algorithms and also by mixing different pore types. "Different pore types would have different properties and therefore different error profiles," he said.
Oxford Nanopore has not tested this, though, he added.
Researchers working on nanopore sequencing technology have tested a number of different types of pores, including alpha-hemolysin and Mycobacterium smegmatis porin A, known as MspA. Additionally, those pores can be tweaked themselves, by inducing mutations at different locations, to give them slightly different properties. Oxford Nanopore has not disclosed what type of pore it uses in its system.
Jaffe did not disclose what the raw error rate of the sequence data was and said that despite the error blocks, the "raw error rate doesn't speak to the utility of the data."
Lex Nederbragt, a bioinformatician with the Norwegian Sequencing Center at the University of Oslo, who attended AGBT, told In Sequence that while the information shown was limited, regarding the error profile, there "seem to be clusters of errors of several bases, and cases with missed bases." It is unclear what the frequency of those errors are and whether there are sequence-dependent biases, he added.
Additionally, "the errors shown limit the applicability of the reads," Nederbragt said. "These reads would be fine for structural variation and RNA isoform detection, but not for de novo genome assembly without the help of short-read data."
Jaffe said that the Broad team used the MinIon data to create better assemblies from Illumina data. The researchers sequenced and assembled the Scardovia genome with Illumina, creating a PCR-free library, sequencing with 250-base reads and assembling with the DISCOVAR algorithm.
The researchers then used the long reads generated from the MinIon to resolve gaps and discrepancies in the Illumina assembly. He demonstrated that the MinIon data could resolve single-base discrepancies in the assembly as well as structural variations and repetitive regions. For instance, in one location, the Illumina assembly indicated that there were either two chromosomes that shared a 6-kilobase repeat or that it was one chromosome with a 6-kilobase repeat. Overlaying the longer MinIon reads showed that it was one chromosome. In another instance, Jaffe showed that the MinIon data was able to resolve the length of a tandem repeat. The Illumina assembly was able to show that the tandem repeat was there, but could not resolve its length.
"At the end, essentially everything was resolved," Jaffe said, with two exceptions—one case where the correct base is unclear and a homopolymer stretch of G bases, where the exact length is not known.
Also this week, Oxford Nanopore began issuing invitations to its MinIon Early Access Program. Participants in the program pay a $1,000 refundable deposit plus a $250 delivery charge. The early access program will consist of two cycles, although participants can choose to drop out after the first approximately six-week cycle and receive their deposit back. The company said it would start shipping the "configuration pack" for the MinIon in the second half of March.
During the first cycle, Oxford will require participants to run a certain number of "burn-in" experiments — control experiments that Oxford Nanopore routinely runs internally. Additionally, the company recommends running the system at a speed of under 30 bases per second per nanopore.
Participants will be asked to report the results of their burn-in experiments to Oxford, after which they can run their own experiments and share results. In cycle 1 of the program, participants will receive four flow cells. After completion of cycle 1, participants can return the flow cells to Oxford and continue into phase 2, or return both the flow cells and the MinIon system and exit the program. Participants that exit will receive their $1,000 deposit.