MinION (courtesy of Oxford Nanopore Technology)

Nicholas Loman of the University of Birmingham has been following the development of nanopore sequencing since Clive Brown, chief technology officer of Oxford Nanopore Technologies (ONT), announced plans for two nanopore sequencers in 2012.

Loman and others were particularly intrigued by the MinION (pronounced 'min-ion'), a handheld sequencer that could be connected to a laptop via a USB port and promised to deliver reads as long as 100 kilobases in real time. In November 2013, ONT started taking applicants for its MinION Access program, in which Loman and his PhD student Joshua Quick were participants.

The MinION consists of a membrane housing around 500 protein pores that connects two chambers filled with electrolytes and allows a current to flow through. As a DNA molecule passes through the pore, the current is blocked, producing a pattern that is characteristic for each base. The challenge is to measure the current with high enough spatial and temporal resolution to decode each base accurately. In June 2014, Loman and Quick posted a read from their first sample run on the MinION to figshare “simply to show that the instrument could do it,” recalls Loman. “Some people thought it was a physical impossibility to sequence DNA with this method.” But he acknowledges that the quality and accuracy of the read were not great.

Part of the problem was a suboptimal library preparation method—two enzymes need to be attached to the DNA to facilitate its processing through the pore. In what ONT calls the R6 version of the library prep, a motor protein binds to one end, a hairpin connects the DNA strands and a second motor protein attaches to the hairpin. The purpose of these proteins is to slow the transition down enough for the individual bases to be recorded. The reads of highest accuracy are '2D reads', in which the template strand is sequenced, followed by the hairpin and the minus strand. Poor binding efficiency of the second motor protein to the hairpin reduced the number of 2D reads in the initial MinION sequencing runs.

In September, ONT presented the R7.3 library prep, with the key improvement of having the second motor protein already ligated to the hairpin. With the R7.3 kit in hand, it took Loman and Quick only two weeks to produce data on an Escherichia coli genome and upload it to the GigaDB sequence database (Quick et al., 2014)1. Loman teamed up with Aaron Quinlan of the University of Virginia to write the program Poretools for quality control and downstream analysis (Loman and Quinlan, 2014)2. He hopes that their data will trigger the further development of tools for better alignment, variant calling and de novo assembly.

Loman also predicts that a mixture of different pores on a flowcell may improve the quality of the reads. Accuracy will also increase with the number of 2D reads. At present Loman sees an accuracy of 85% if both strands run through the pore correctly. If only one strand gets through, the accuracy drops to ∼70%.

Loman deems ONT's handheld device as ideally suited for diagnostics or environmental monitoring that does not rely on very high accuracy. He is also leading efforts to field test the device.