AGBT 2016 is happening now and it has been two years since we heard about the first MinION experiment at AGBT 2014. It has been two years since Oxford Nanopore selected participants for its MinION early access program. And Oxford Nanopore has come a long way since the last two years.
Two very interesting papers showing the real time application of Oxford Nanopore technologies came out last week. The first one published on Nature showed the application of MinION for Ebola surveillance.
- Real-time, portable genome sequencing for Ebola surveillance, Joshua Quick, Nicholas J. Loman, Sophie Duraffour, Jared T. Simpson, Ettore Severi, Lauren Cowley, et. al.
The second paper published on bioRxiv, showed for the first time the demonstration of selective/targeted real-time sequencing using Oxford Nanopore’s Read Until technology.
Read Until/Selective Sequencing by Oxford Nanopore MinION
At ASHG 2014, at the special Oxford Nanopore event, Oxford Nanopore hinted the basic idea behind “Selective” sequencing in real time. Basically, if one is interested in sequencing specific sequence of bacteria/virus in a sample, (the current approach of) sequencing everything in the sample and then finding out if the individual is infected or not is probably not the best approach.
Oxford Nanopore said that it can allow “selective” or targeted sequencing by letting one to monitor what is being sequenced in each pore and give users the option whether to continue sequencing only the sequence of interest. If it is from the region of interest, you let the nanopore finish sequencing the read. If it is not from the region of interest, a user can stop the sequencing.
The new sequencing paradigm for real time analysis was a very a tantalizing idea. One can easily see the number of potential applications of such technology. Later in summer of 2015, Oxford Nanopore named the selective sequencing as “Read Until”.
Selective sequencing aka “Read Until” is not just a “tantalizing idea” anymore. Matthew Loose et al from University of Nottingham have beautifully demonstrated the selective sequencing idea in the new bioRxiv preprint. The authors demonstrated selective sequencing by using bacteriophage lambda DNA. The small phage genome with 48,502 bp was ideal this demonstration.
The way it works in Nanopore technology is that, just as the reads go through the pore, MinION can stream current measurements, known as squiggles, from all the pores in real time. The value of current at a time point is determined by the specific DNA base that is in contact with the nanopore.
Oxford Nanopore offers Read Until API that lets users control which pore should selectively sequence. Each Nanopore channel is also addressable individually and one can stop the sequencing reaction in a pore by “simply” reversing the potential difference.
Once we stop the sequencing of the DNA molecule, the Nanopore let that molecule go and it can sequence the next available DNA molecule.
Dynamic Time Warping algorithm for mapping read squiggle to genome squiggle
One of the challenges in making the selective sequencing work is figuring out the bases of the current squiggle in real-time and deciding whether to continue sequencing or not. Naive approach would be to do base calling on the sequenced fragments from each pore and decide to sequence or not. However, current base calling approach is too slow for making it happen in real-time.
Instead of base calling, the authors worked on mapping directly squiggle data from each pore on to squiggle data corresponding to a reference sequence using a dynamic time warping (DTW) algorithm. DTW is a dynamic programming algorithm for matching time series data that may differ in phase and amplitude and it has been used in speech recognition first and on other time series data later. The main rationale behind DTW is that in comparing time-series data like squiggle, the standard distance measure (like euclidian distance) of time points from one data to another is not the right measure of similarity.
In the first experiment, Loose et al selectively sequenced two 5kb regions of the genome, from 10-15 kb and 30-35 kb and rejecting all other reads. As a nice control, the authors applied ‘Read Until’ only on even numbered channels. In the figure above, one can clearly that read depth from odd channels is constant across the genome with 500X coverage. In the even channels, there are two nice bumps in the coverage at genomic locations 10-15k and 30-35k corresponding to the regions selected for sequencing. IN the second experiment, the authors showed selective sequencing using amplicons. With real-time selective sequencing, fascinating times ahead indeed.
Read Until selective sequencing was also in action at AGBT 2016.
— mattloose (@mattloose) February 12, 2016