More QuantumSi Pondering
Over on the Discord we’ve been pondering upon QuantumSi’s new non-chip approach.
I thought it might be interesting to summarise here. Feel free to comment below, or add your thoughts on the Discord!
The big over-arching question is:
What’s the point in moving away from a chip, if you have to monitor in real-time anyway?
I’ve discussed this before. But the cleavage point explicitly seems worth revisiting.
In terms of explicit statements from QuantumSi, as far as I can tell this is kind of all we have to go on:
“Proteus design and planned capabilities leverage existing proven technologies already in place at the Company, including consumable surface chemistry, sequencing chemistry, recognition technology and optical technology”
My stance is that they probably need to drop the real-time aspect. That is, the real-time cleavage approach.
Option 1 - Monitor Array In Realtime
The chips previously shown look like they might have 4 imaging regions. So you could imagine that they could just run each region as its own sequencing experiment. Runs currently take ~10 hours. So, this would push you to 40 hour run times.
Scaling to more regions, would mean scaling run time1.
Option 2 - Cyclic Cleavage
Alternatively they could move to cyclic cleavage. That’s going to require a fluidic system to introduce a cleavage reagent periodically.
There might be something fancy they can do here to keep their enzymatic cleavage approach, I can imagine an SBB type approach where they bind enzymes in one step, free enzyme is flushed and then is convinced to remove a single AA and quickly flushed.
But they could also just use Edman degradation.
Here they would monitor each region for a fixed period of time, accumulating binding kinetics information. Then once this is complete, run a cycle of cleavage.
Because they are probably dropping life time2 for a color ratio output, they can likely just accumulate binding information for a fixed period of time.
It’s tempting to consider that they might be able to drop the realtime accumulation entirely and just use color ratios. That would open up imaging techniques like TDI, which would allow fast imaging cycles across large arrays.
Summary
See my previous notes for more thoughts. But if they continue to monitor the whole array in realtime on the off-chip system then the advantage of moving off-chip seems much more limited.
A cyclic approach brings a lot of complexity (fluidics, cleavage chemistry) but might give them an easier route to pushing throughput.
Thoughts? Why not share on the Discord!
Or scaling number of optical systems in the box, like the NextSeq 550. Or moving to bigger cameras.
Slides previously shown strongly suggested this:
Any idea what the structure in the top right corner of the image of the 20M flowcell is? Fluidics ports?
Also, if using cyclic chemistry, they'd presumably lose the kinetic info that they relied upon to distinguish between similar amino acids like Leu vs Ile vs Val, so they may need to compensate for that..