Ciro Santilli @cirosantilli 40

Their main innovation seems to be their 3D design which they call "Coaxmon".

Funding:

The normal navigation to them was paywalled, but the static files are served without login checks if you know their URL. One way to go about it is to search by prefix on the Wayback Machine: web.archive.org/web/*/https://www2.physics.ox.ac.uk/sites/default/files/contentblock/2011/06/03/*

The last handbooks we can find are 2020/2021, they might have move to a new more properly paywalled location after that year.

At the time of the experiment, Illumina equipment was cheaper per base pair and dominates the human genome sequencing market, but it required a much higher initial investment for the equipment (TODO how much).

The reusable Nanopore device costs just about 500 dollars, and about 500 dollars (50 unit volume) for the single usage flow cell which can decode up to 30 billion base pairs, which is about 10 human genomes 1x! Note that 1x is basically useless for one of the most important of all applications of sequencing: detection of single-nucleotide polymorphisms, since the error rate would be too high to base clinical decisions on.

Compare that to Illumina which is currently doing about an 1000 dollar human genome at 30x, and a bit less errors per base pair (TODO how much).

Other advantages of the MinION over Illumina which didn't really matter to this particular experiment are:

Now that's some basal shit! It's basically a fucking blob!!! Except that it is flat. No nervous system. Not even tissues. It is basically a multicellular

Inferior compared to self-directed learning, but better than the traditional "everyone gets the same" approach.

With all this ready, we opened the Nanopore flow cell, which is the 500 dollar consumable piece that goes in the sequencer.

We then had to pipette the final golden Eppendorf into the flow cell. My anxiety levels were going through the roof: Figure 4. "Oxford nanopore MinION flow cell pipette loading.".

At this point bio people start telling lab horror stories of expensive solutions being spilled and people having to recover them from fridge walls, or of how people threw away golden Eppendorfs and had to pick them out of trash bins with hundreds of others looking exactly the same etc. (but also how some discoveries were made like this). This reminded Ciro of: youtu.be/89UNPdNtOoE?t=919 Alfred Maddock's plutonium spill horror story.

Luckily this time, it worked out!

We then just had to connect the MinION to the computer, and wait for 2 days.

During this time, the DNA would be sucked through the pores.

As can be seen from Video 1. "Oxford Nanopore MinION software channels pannel on Mac." the software tells us which pores are still working.

Pores go bad sooner or later randomly, until there are none left, at which point we can stop the process and throw the flow cell away.

48 hours was expected to be a reasonable time until all pores went bad, and so we called it a day, and waited for an email from the PuntSeq team telling us how things went.

We reached a yield of 16 billion base pairs out of the 30Gbp nominal maximum, which the bio people said was not bad.

www.qiagen.com/us/products/discovery-translational-research/dna-rn-a-purification/dna-purification/dna-clean-up/qiaquick-pcr-purification-kit/#orderinginformation (archive)

Manual archive: web.archive.org/web/20190911100243/https://www.qiagen.com/us/resources/download.aspx?id=e0fab087-ea52-4c16-b79f-c224bf760c39&lang=en

Removes PCR byproducts from purified DNA.

Biology experiments are hard, and so they go wrong, a lot.

For this reason, it is wise to verify that certain steps are correct whenever possible.

And so this is the first thing we did on the second day!

Gel electrophoresis separates molecules by their charge-to-mass ratio. It is one of those ultra common lab procedures!

This allows us to determine how long are the DNA fragments present in our solution.

Since we know that we amplified the 16S regions which we know the rough size of (there might be a bit of variability across species, but not that much), we were expecting to see a big band at that size.

And that is exactly what we saw!

First we had to prepare the gel, put the gel comb, and pipette the samples into wells present in the gel:

To see the DNA, we added ethidium bromide to the samples, which is a substance that that both binds to DNA and is fluorescent.

Because it interacts heavily with DNA, ethidium bromide is a mutagen, and the biology people sure did treat the dedicated electrophoresis bench area with respect! Figure 4. "Gel electrophoresis dedicated bench area to prevent ethidium bromide contamination.".

The UV transilluminator we used to shoot UV light into the gel was the Fisher Scientific UVP LM-26E Benchtop 2UV Transilluminator. The fluorescent substance then emitted a light we can see.

As barely seen at Figure 8. "Fischer Scientific UVP LM-26E Benchtop 2UV Transilluminator illuminated gel." due to bad photo quality due to lack of light, there is one strong green line, which compared to the ladder matches our expected 16S length. What we saw it with the naked eyes was very clear however.

More generic PCR information at: Section "Polymerase chain reaction".

Because it is considered the less interesting step, and because it takes quite some time, this step was done by the event organizers between the two event days, so participants did not get to take many photos.

PCR protocols are very standard it seems, all that biologists need to know to reproduce is the time and temperature of each step.

We did 35 cycles of:

This process used a Marshal Scientific MJ Research PTC-200 Thermal Cycler:

We added PCR primers for regions that surround the 16S DNA. The primers are just bought from a vendor, and we used well known regions are called 27F and 1492R. Here is a paper that analyzes other choices: academic.oup.com/femsle/article/221/2/299/630719 (archive) "Evaluation of primers and PCR conditions for the analysis of 16S rRNA genes from a natural environment" by Yuichi Hongoh, Hiroe Yuzawa, Moriya Ohkuma, Toshiaki Kudo (2003)

One cool thing about the PCR is that we can also add a known barcode at the end of each primer as shown at Code 1. "PCR diagram".

This means that we bought a few different versions of our 27F/1492R primers, each with a different small DNA tag attached directly to them in addition to the matching sequence.

This way, we were able to:

Input: Bacterial DNA (a little bit)
... --- 27S --- 16S --- 1492R --- ...

|||
|||
vvv

Output: PCR output (a lot of)
Barcode --- 27S --- 16S --- 1492R

Finally, after purification, we used the Qiagen QIAquick PCR Purification Kit protocol to purify the generated from unwanted PCR byproducts.

www.gelifesciences.com/en/us/shop/protein-analysis/protein-sample-preparation/protein-enrichment/magrack-6-p-05761 (archive).

www.bidspotter.com/en-us/auction-catalogues/bscsur/catalogue-id-bscsur10011/lot-c6605b41-1a14-40e5-a255-a5c5000866e0 (archive) Cannot exact same product on official website, but here is a similar one: www.fishersci.co.uk/shop/products/lm-26-2uv-transilluminator/12382038 (archive).

PuntSeq is a side project led by a few University of Cambridge PhDs that aims to determine which bacteria are present in the River Cam.

In July 2019, the PuntSeq team got together with the awesome Cambridge Biomakespace, an awesome biology makerspace open to all, to create a two day science outreach activity showing their procedures.

The data collected in this experiment, together with other collection sessions done by the organizers actually led to a publication on eLife: elifesciences.org/articles/61504 "Freshwater monitoring by nanopore sequencing" by Lara Urban et al. (2021), so it is awesome to see that were are actual being part of "real science".

Ciro knows nothing about biology, but since he is very curious about it, he jumped at this opportunity, and decided to document things as well as his limited knowledge would allow.

All participants chipped in some money to help cover the experiment's costs. Ciro suspects that this activity was done partially to help crowdfund the experiment, but it was a worthy investment!

The impressions you get from the experiment as a software engineer will be:

Because Ciro's a software engineer, and he's done enough staring in computers for a lifetime already, and he believes in the power of Git, he didn't pay much attention to this part ;-)

According to the eLife paper, the code appears to have been uploaded to: github.com/d-j-k/puntseq. TODO at least mention the key algorithms used more precisely.

Ciro can however see that it does present interesting problems!

Because it was necessary to wait for 2 days to get our data, the workshop first reused sample data from previous collections done earlier in the year to illustrate the software.

First there is some signal processing/machine learning required to do the base calling, which is not trivial in the Oxford Nanopore, since neighbouring bases can affect the signal of each other. This is mostly handled by Oxford Nanopore itself, or by hardcore programmers in the field however.

After the base calling was done, the data was analyzed using computer programs that match the sequenced 16S sequences to a database of known sequenced species.

This is of course not just a simple direct string matching problem, since like any in experiment, the DNA reads have some errors, so the program has to find the best match even though it is not exact.

The PuntSeq team would later upload the data to well known open databases so that it will be preserved forever! When ready, a link to the data would be uploaded to: www.puntseq.co.uk/data

One of the sequencers made by Oxford Nanopore Technologies.

The device has had several updates since however, notably of the pore proteins which are present in the critical flow cell consumable.

Official documentation: nanoporetech.com/products/minion (archive)

The following images of the device and its peripherals were taken during the experiment: Section "How to use an Oxford Nanopore MinION to extract DNA from river water and determine which bacteria live in it".

Moodle instance of the Mathematical Institute of the University of Oxford.

Has a mixture of open access and closed access. But at least it can have open access unlike the in-house systems such as Canvas where everything is necessarily paywalled!

Sometimes things appear open but don't show any meaningful content if you are not logged in, which is annoying.

But at least it gives a clear public course list, thing that certain departments (cough Department of Physics of the University of Oxford cough).

The organization is a bit crap, when you expand e.g. C Michaelmas term it shows nothing, just a search.

The way to go is via the year year categories e.g. "Year 2022-23": courses.maths.ox.ac.uk/course/index.php?categoryid=734. Term splitting is annoying, but one can stand it.

There seems to be no way to list all versions of a single course across multiple years besides just doing a search e.g.

This job announcement from 2022 gives a good idea about their tech stack: web.archive.org/web/20220920114810/https://oxfordionics.bamboohr.com/jobs/view.php?id=32&source=aWQ9MTA%3D. Notably, they use ARTIQ.

Funding: