Saturday, August 20, 2016

Where the bleep did my pores go??

Another afternoon spent this week in the lab, and another awful MinION run.  I figured that the last run was eminently beatable, given that I ended up with exactly one read aligning to my reference (though perversely, re-calling the data via Metrichor ended up with two reads in the pass folder -- one of which had zero bases in it -- I guess it is a passing null read).  But no, despite coming close to being botch-free on the routine lab stuff, dark clouds came early and the experiment yielded not a single usable read.

So a new package arrived from ONT, with wash kits and a flowcell of the new SpotON type.  The wash kits were used to wash the previous three flowcells, two of which quickly failed platform QC and the third I terminated when I saw a huge bubble over the sensor array at the start.  So on to the new flowcell.

MinKNOW, of course, required an update, expected because new scripts for the SpotON flowcells.  This started to look Sisyphean, as I cycled multiple times between starting MinKNOW, being told it needed to update, updating, rebooting and then in the loop again.  After 2-4 cycles, somehow I exited.  

Platform QC looked mostly promising -- there was clearly a region in the lower right corner of the right-hand segment in the MinKNOW display that had no good colors on any mux scan, but overall there were something like 400 active pores -- though it was hard to tell because MinKNOW leaves that display once the QC is done.  Again, I do not like this; I want to queue up the entire process at the start and then have the software pause.

So now to priming and a crucial decision.  The SpotON devices have a new twist.  They can be loaded like the old flowcells, but alternatively there is a tiny loading port essentially over the sensor array, allowing loading with far less dead volume.  This is also going to key for future advances such as VolTRAX and Zumbador, which will deposit beads via this port.  

The port is covered by a truly minuscule (I don't have it in front of me; in my memory it's about the size of an mid-size ant)  "activator", which I would later lose track of (but find again) -- I definitely suggest that one hold onto some spares, as their size suggests they will invariably go missing.  The priming process is slightly different -- first 500ul with the activator in place, then 300ul again with it in place and finally 200ul with the activator removed.

Now, at this point a decision.  Do I stick with the existing procedure of try the new?  On the one hand, the new is slightly more complicated, so more to go wrong.  On the other, why lock into an obsolete procedure? Of course, a little horned version of me carrying a three-channel piper tie is sitting on my shoulder and convinces me to go the new route 

As far as I could tell, the first two primings went well -- I successfully aspirated the small bubble at the loading port and then pushed in the buffer, with no visible bubbles.  Well, on the first try on the second priming nothing happened - somehow I had the piper tip misaligned with the port.

So now I remove the activator "bung" and push 200ul of buffer - and a big puddle emerges from the bunghole. NOBODY SAID ANYTHING ABOUT A PUDDLE! Can't be good - so I wick it away with a kimwipe.

So now to prepare the loading mix, which apparently was my cue for a self-sabotage attempt. 20ul filter tips fit a P200 but are good for pie tying 35ul! 

Then, the new loading procedure, which calls for dripping the loading into the bunghole, letting each drop be slurped up before adding another. Which turns out to not be easy. This is where I could use a modified tip with a bung and activator on the side so I could break the seal so it just drips slowly. 

Now I start the running script, and the calibration fired up. 40 active pores. Somehow, despite not seeing an air bubble in the system, I've killed 90% of the delivered flowcell performance. AIEEEEEEE!!!

Roughly simultaneously with my floundering, PoreCamp was taking place at a beach area in England, with much nicer results. Apparently one run yielded 3Gb, and most if not all runs were quite healthy. And Clive Brown is advertising the next version the the flowcells will have twice the output. 

ONT needs to figure how to franchise this approach, or they risk having a large number of initial users giving up quickly. I can personally imagine running s few flowcells a year as a hobby (still far cheaper than that of a buddy down the street who drag races). But not if it's just an endless string of busts at nearly $1K each.  Back in the 70's my father and oldest brother built a kit computer and it wasn't shocking in those days for a capacitor to die in a puff of smoke, But losintg the CPU on every run might of quickly dampened their enthusiasm.  If ONT has batches of substandard flowcells, perhaps they should market them cut-rate as learners kits.

At last year's London Calling ONT discussed "crumpet" chips which would separate the liquid channels from the pricey ASIC electronics.  ONT has proposed pricing this technology, at least as an option, simply by the amount of data generated. I could really go for this - transferring the financial pain of my misadventure from me to ONT!

An awful cliche says those who can't do will teach, so I may just start instructing our skilled molecular biologists in what I know but cannot execute.  What could go wrong?

P.S. - if you'd like to see a much better first experience with a MinION, check out Lisa Cohen's blogpost on using them at the Molecular Diversity course given at Marine Biology Lab in Woods Hole


Anonymous said...

Keith, I think your tone is harsh. Keep in mind that this is putting sequencing into the hand of the masses, taking what is/was a complicated process and making it so simple that high school students are now using it.

Perspective and context are powerful.

Keith Robison said...

Fair charge - but I would argue that as sequencing is brought to the masses there will be many well-meaning klutzes like me who will experience this degree of frustration. There may be simple fixes - on Twitter it was suggested my pipetting may be too abrupt, which might be solvable with practice

Anonymous said...

If a PhD with prior lab experience is having issues, I would find it rather hard to believe any high school students are actually using it and getting meaningful results.
Perspective and context are powerful, as they were with the car that ran on H20, cold fusion and jet packs.

Anonymous said...

You don't have to believe, you just have to read....

Jonathan Jacobs said...

Our experiences with the MinION have also been underwhelming - but hey, they can sequence in space now. ;) Snark aside - we'll keep trying - even though it is pretty expensive to fail. It's not surprising that the tech is still part Art and part Science. Last time we ran a MinION a couple weeks ago I remember thinking: this seems eerily similar to those early Northern Blots i did in grad school. The CSH protocols never captured all the details, and each gel, transfer, exposure, and film we developed was a unique experience.

James@cancer said...

Hi Keith, I too have had issues getting the R7s to do anything other than generate junk. However the R9s are working and we're now getting data on a real project. The reality is that we hear a lot more about the successes from Loman et al, and the sequencing in space, but we don't hear too much about the failures. I'm still looking forward to one ONT development that has not been discussed and that is truly open access to the technology i.e. moving away from the closed MAP. For now I'm happy to be generating what is really exciting data and developing a method that would be impossible on alternative platforms.
I can't wait to get my hands on SmidgION. And am really looking forward to the RNA developments.
Have fun...maybe let your son have a go?

Mark said...

With respect of the spot on cells loading - it can be done, but practice a bit first (on a used flowcell). BTW: The liquid handling experience learned from refilling various types of inkjet cartridges is also very useful when working with nanopore:

To load the library on the spot on:
1. Open priming lever;
2. Draw a bit of liquid (2-3ul) from priming port;
3. Remove spot on plug;
4. draw 1-2 ul from the plug hole, and drop it on the top - so the fluid is right to the top of the priming hole;
5. Now put a few drops of the priming buffer near the hole, wait for it to sink in;
6. if it does not want to sink in - make a liquid bridge to the hole, draw 1 ul to remove the bubble trapped in the hole;
7. Once done with priming - now you can add library;
8. If possible use a wide bore tip to load a library - just place drops on a ramp nest to hole, and wait for them to sink in;
9. Once finished, first plug the spot on hole;
10. Do not forget to close the priming lever.

Anonymous said...

ONT so farm from our experience is the opposite of illumina in every way... unfortunately this includes ease of use and risk of costly errors. Does anyone know if promethion loading is the same level of difficulty? I couldn't imagine loading 48 fc every 2 days like this.