41J Blog

Below is a list of DNA Synthesis Companies, to complement my list of sequencing companies. It’s not quite as complete, I’ve missed out some seemingly established players who didn’t seem particularly entertaining and/or only run service businesses.

There’s a great list here which includes some defunct companies, and other approaches.

Northshore Biosciences popped up on my radar again recently. There’s not a lot of information on the web, so I decided to skip forward in my list of DNA sequencing companies and write up a few notes on them.

Business

Northshore Biosciences was founded in 2009 as Lux Bio Group, Inc. [8] by Jonathan DeHart and Gordon Holt. A 2013 Genomeweb article states that an undisclosed series A was raised from Oregon Angel Fund and the ISB (Institute for Systems Biology?) [2].

A 2014 report by the Keiretsu Forum (a group of angel investors) states that they invested 21.2M USD in 35 companies during 2013 (including Northshore Bio). The average investment from Keiretsu was ~600K. SEC filings seem to indicate they’ve raise about 3M USD. Overall, it appears they’ve raise a few million and are at series A stage.

Technology

The Northshore Bio site doesn’t describe the technology in any depth, but it does show a nice video.

The fundamental Northshore approach is to create what they call “tuneable nanopores”. This is a fabrication approach where they create an aperture and then try and fill it in until they have a much smaller hole through which they can detect the translocation of bases. The genomeweb article suggests they are targeting 20nm long and 10nm wide nanopores.

It appears they are also proposing what they call “sequencing-by-degradation”. This is similar to the original Oxford Nanopore approach. Here an exonuclease, positioned near the aperture chews individual bases off a strand. These individual bases then go through the pore and are detected.

The pore size suggested (20nm by 10nm) is large when compared to typical protein nanopores, which typically have constrictions of 1 to 3nm:

Example Protein Nanopore dimensions from [5].

Solid state nanopores too, have achieved dimensions in the single digit nanometer range:

Example solid state nanopores from [6].

So it’s likely that the large dimensions of the pore somewhat motivate the decision to use an exonuclease, single nucleotide detection approach. One potential issue here is the dwell time of the bases in the nanopore. In many single base detection experiments the dwell time of bases is non-gaussian. There are many nucleotides that go through the pore so quickly that you can’t detect them:

Base dwell times from [7].

In addition to the basic method shown on their site, Northshore Bio appear to have a number of patents (still assigned to Lux Bio Group Inc.) [1].

The patents discuss a number of different approaches. These include:

* Creating a small aperture, in which a bilayer just big enough for a single nanopore is formed (this could help with issues with multiple insertion of pores in other systems).
* Nanowells with sidewall electrodes.
* Nano-membranes which can change shape.
* Depositing membranes under electro-chemical control.

Most of the patents appeared to be continuations, and all authored by Gordon Holt. I looked through the patents for data (SEM images, or experimental data) and couldn’t find anything. It’s possible there’s more stuff in the pipeline.

The Northshore approach seems interesting, but not without significant challenges. I’d guess they will need significant funding to move forward (anything that involves nano fabrication does!). Will be watching with interest!

Notes

[1] Patent, Lux Bio Group: http://www.freepatentsonline.com/y2017/0298432.html
[2] Genomeweb article: https://www.genomeweb.com/sequencing/northshore-bio-develops-solid-state-tunable-nanopore-chips-sequencing-degradatio
“In late 2011, NorthShore Bio raised an undisclosed amount of funding in a Series A round with the Oregon Angel Fund and the ISB.”
“For sequencing applications, the company is targeting pore dimensions of less than 20 nanometers in length and 10 nanometers in diameter, similar to the nanopores explored for sequencing by others.”
“For sequencing, NSB is pursuing a sequencing-by-degradation approach, which is similar in principle to the exonuclease sequencing strategy Oxford Nanopore was exploring before it abandoned it in favor of DNA strand sequencing.”

[4] https://www.k4northwest.com/down/eJzLKCkpsNLXL87MyS4uSSwq0Ss21kvMTazKz0ssL9ZLzs%40VNzU2TjMyNDcB0pYG5gYphhYmZkaJpsZ6BSlpAJ08E30%3D/Keiretsu%20%20Forum%20Northwest%202013%20Funding%20Press%20Release.pdf

[5] https://doi.org/10.1016/j.tibtech.2011.07.006

[6] https://www.researchgate.net/publication/303696326_Solid-State_Nanopore-Based_DNA_Sequencing_Technology

[7] http://www.nature.com/articles/nnano.2009.12

[8] There’s still an old website online for Lux Bio Group: http://luxbiogroup.com/index.html

[9] https://www.whoisraisingmoney.com/lux-bio-group-inc

Business

Direct Genomics was founded in 2014 [2] in Shenzhen. Their approach uses IP first explored at US DNA sequencing company Helicos (which went bankrupt in 2012). The website states that they have raised ~29M USD [1]. Other sources suggest that Cosun Venture Capital have invested ~34M USD in April 2018. I can find no other investments by Cosun Venture Capital. I’ve not seen references to other investors (aside from some government grant funding).

Technology

The Direct Genomics is, by all accounts, a reboot of the Helicos technology. Many of the original Helicos sequencers are now owned by SeqLL who keep them running as a service business. They have a nice video on their site which describes the platform which seems largely identical to what is proposed by Direct Genomics.

In the Helicos approach a flow cell is covered with anchored polyT single stranded DNA. The fragments you want to sequence are prepared with a polyA tail and hybridise with the polyA fragments. This gives you a flowcell covered with single stranded DNA.

While the specifics are different, this isn’t hugely different from what Illumina do (circa GA2->Hiseq 2000). The fragments are randomly attached to the surface.

On the Illumina platform these single molecule would then undergo amplification (cluster growth). This duplicates the single strand, creating a small cluster of identical fragments. On the Helicos/Direct Genomics platform, we just have a single strand (single molecule).

The Direct Genomics platform then sequences the strands using a sequencing-by-synthesis method. Assuming the chemistry hasn’t been upgraded since the Helicos, they still use a single dye. This means that rather than being able to flow in all 4 bases at once, they need flow them in one at a time. The Helicos virtual terminator technology performs the same purpose as the Illumina reversible terminators, and prevents multiple incorporations.

Why they don’t use 4 different dyes is a bit of an open question, some posts suggest potential IP issues [6]. I wonder if the terminators are not that efficient, and significant potential for multiple incorporation remains, incorporating a single base at a time might help limit this.

The Helicos/Direct Genomics technology is almost a single molecule implementation of the Solexa approach (with of course significant chemistry differences). It’s relatively well known that Solexa also tried single molecule approaches [5] before abandoning them because clusters worked so much better.

The Helicos single molecule approach worked (but only generated very short reads of 25+bps). What exactly limited read length is unclear to me. . It seems likely that continued illumination might eventually cause templates to “fall off” the flowcell or cause other damage. Illumina reads also fade with read length so the issue isn’t unique to Direct/Helicos. But with a single molecule SBS approach like this, any damaged caused will terminate sequencing of this fragment.

Single molecule imaging has made significant improvements in recent years. Direct Genomics has this advantage in their favour. One report suggests they’ve been able to achieve 200bp reads [2] though no word on error rates.

Their’s also one other key difference with Direct. They don’t appear to be going after the whole genome sequencing market. Rather than using polyT sequences on the flowcell, Direct design probes to target particular regions on interest in the genome. This is clearly aimed at clinical applications, where they can create flow cells that target mutations of interest.

They’ve certainly raised a bunch of cash, and it will be interesting to see how things play out this time.

Notes

[1]
http://www.directgenomics.com/index.php/portal/js_article/newsdetail/id/34
“This product is the result of five years of hard work by the science team at Direct Genomics, which had almost gone bankrupt twice due to shortages in funds,” said He, who tapped into the upstream sector of the sequencing industry in 2012 after finishing his postdoctoral training at Stanford University.

However, thanks to the courage of venture capitalists and the generosity of the Shenzhen government in supporting startups, the company, which started with 1 million yuan (ed: ~150K USD), has raised 200 million yuan (ed: ~29M USD) for its development in recent years. The city provided 40 million yuan (ed: ~6M USD) in subsidies under the Peacock Program for the development of the company.

[2]
BioIT World article: http://www.bio-itworld.com/2015/10/29/direct-genomics-new-clinical-sequencer-revives-forgotten-dna-technology.html

[3] https://www.chinamoneynetwork.com/2018/04/19/dealshot-sequoia-capital-leads-48m-series-c-round-in-baby-products-shopping-platform-patpat

[4] Virtual Terminator paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2719685/

[5] Solexa originally single molecule: https://core-genomics.blogspot.com/2012/08/what-happened-to-illuminas-single.html

[6] http://seqanswers.com/forums/archive/index.php/t-20466.html

Business

Depixus (originally Picoseq) was founded in 2012 in Paris, based on IP developed by Vincent Croquette, David Bensimon, and Jean-François Allemand at École Normale Supérieure (ENS). In 2013 the company was seeking 5M Euros, but it’s not clear to me if the round closed [1]. In May 2016 Arix Biosciences agreed to acquire up to ~20% of the company for 1.24M Euro (paid 0.93M Euro, with it appears the rest relying on milestone payments, though reports vary) [2].

While the company is based in Paris, a Cambridge (UK) branch exists. Full accounts appear to be filed in the UK [3] and it seems that at the end of 2016 they had around 1M Euro is cash, and 800K Euro in cash to come. In 2016 operating costs were listed at 1.3M Euro, personnel costs at ~600K Euro.

Technology

The technology behind Depixus is described in a 2012 paper. Fundamentally the detection process relies on the preparation of hairpin’d double stranded DNA (double stranded DNA with a loop on the end). One end of this strand is connected to a surface, the other to a magnetic bead.

The bead is held in a magnetic field that can be varied. By varying the field the bead can be moved, pulling the DNA. You can therefore pull the double stranded DNA apart “unzipping” it. It’s also possible to reduce the field strength, and allow the DNA to zip back up. While applying this force, you can also sense the location of the bead. In the 2012 paper this is performed optically, by analysing the diffraction rings around the illuminated beads:

This system allows Depixus to measure the forces required to zip/unzip the DNA, and it is this measurement, and interference with the zipping process, that Depixus uses to sequence DNA.

Ligation based sequencing from 2012 paper.

In the 2012 paper [5] they show a hybridisation and a ligation scheme (illustrated in the schematic above). In this scheme they fully unzip the strand then ligate a probe (7mers, all Ns expect for a single known location). They then allow the strand to zip up again. If ligation was successful, the bead position is offset by ~5nm. This change in position, tells you if the probe ligated, and therefore which base is present at the known location. They then cleave off 6 nucleotides (if ligation was succesful) and perform the process cyclicly to determine the complete sequence of the strand.

I briefly looked for additional progress on the platform since the 2012 paper. The website currently suggests that they are able to detection base modifications, but I was unable to determine the exact mechanism. One report on their Arix deal, suggests they have generated 286bp reads [2], but again provides few details.

A patent from 2016 [4] covers an improved detection system. Rather than imaging the beads directly they sit the beads in a well. The solution is conductive (contains ions I guess), and there is a bias voltage from the bottom of the well, into the solution. As the bead moves up and down it blocks the aperture of the well. This blocks the ionic current flow, effectively varying the resistance. So, you can measure that current blockage at determine how much the bead is blocking the aperture, and therefore its location.

I’ve not yet been able to find other patents covering the detection of modified bases, it’s possible that they’re not yet available or that they can be found with further investigation.

The technology is quite interesting, in some ways it reminds me of Cearus (detection via changes in beads location/movement) and Complete (ligation). As always it will be interesting to see how things progress.

Notes

[1] Genomeweb article. https://www.genomeweb.com/sequencing/picoseq-expanding-potential-applications-hairpin-dna-based-technology-it-enters
“The company is currently seeking funds from a variety of sources and/or potential partners, PicoSeq CEO Gordon Hamilton said, and hopes to reach its current target of roughly €5 million ($6.6 million) late this spring or sometime over the summer.”
…
“Even so, Hamilton emphasized that he does not see PicoSeq as a direct competitor to existing or anticipated next-generation sequencing instruments. Rather, he argued that the mechanical hairpin method might have a range of applications that are distinct from — or complementary to — those offered by such platforms.

For instance, because each DNA hairpin can be opened and closed tens of thousands of times, it should be possible to look at the same molecule multiple times — perhaps searching for a DNA fingerprint or barcode in one instance and profiling a full sequence or a suite of specific epigenetic modifications in another.

“You could potentially build up the level of detail that you want from the picture based on your requirements as a user,” Hamilton said, “rather than taking the kind of sequencing approach where you chuck everything in and sequence the whole lot and deal with the data afterwards.”

[2] Various references to Arix Biosciences investment.
https://propertibazar.com/article/prospectus-arix-bioscience_5a4f5e8fd64ab2089e446872.html
“On 6 May 2016, Arix Holdings entered into an agreement with Depixus pursuant to which it acquired an interest of approximately 21 per cent. of Depixus’ share capital for an aggregate amount of €1.24 million, paid in three tranches upon achievement of certain technological development milestones in May 2016, December 2016 and July 2017.”

https://www.researchpool.com/download/?report_id=1284000&show_pdf_data=true
“Depixus is a laboratory technology company that is developing novel methods of sequencing DNA and RNA that reveal epigenetic information in addition to the nucleotide sequence. In May 2016, Arix entered into an agreement to acquire up to 18% of the company for €1.24m, of which €0.93m has been paid (for an assumed current stake of 13.5%).”…”The technique is performed by attaching a small magnetic bead to a hairpin of DNA immobilised on a solid substrate and then using magnetic force to unfold the hairpin. A soluble oligonucleotide is added to the solution and the strand is allowed to refold. If the oligonucleotide binds to the DNA strand, it stalls the folding process and alters the force between the bead and magnet. Using a series of oligonucleotides in this way, a complete sequence can be determined. The longest complete sequencing read that has been reported was 286 base pairs, but much longer sequences (over 20,000 base pairs) have been interrogated using the technique, posing the possibility of significant read lengths.”

Arix Bioscience invested in Depixus:
https: //www.researchpool.com/download/?report_id=1284000&show_pdf_data=true

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=26&ved=2ahUKEwi-1Irst8TcAhVLurwKHXh2CWE4FBAWMAV6BAgFEAI&url=https%3A%2F%2Fwww.gov.uk%2Fgovernment%2Fuploads%2Fsystem%2Fuploads%2Fattachment_data%2Ffile%2F725229%2FInnovate_UK_funded_projects_from_2004_to_1_July_2018.xlsx&usg=AOvVaw3FUMKtIp67PHBWQojQTJ4E

[3] https://beta.companieshouse.gov.uk/company/FC032917/filing-history
Accounts as of end of 2016, appear to have had 1M Euro in cash in hand, and 800K Euro to come.

[4] Patent from 2016. It’s assigned to the research institution and Depixus CEO rather than Depixus directly, which seemed a bit weird to me. But it appears to be associated with the Depixus technology. https://patents.google.com/patent/WO2016177869A1/en?oq=Gordon+Hamilton+sequencing

[5] 2012 paper: https://www.nature.com/articles/nmeth.1925