Hi All. I’ve been really struggling to get some precise and accurate results for a particular volume range I need to work with.
For background, we are diluting plasma 1:10 in a 50% Glycerol/PBS solution (manually). The total volume for this solution is 100 uL. We have a Tecan Fluent 780 where we are trying to dilute these samples further into our diluent solution (1X PBS,.05% Tween 20, 1% BSA) via a worklist. I’ve been able to use the adjust accuracy function on the Fluent to dial in the diluent across the volume ranges that we need for diluent (determined by empirically measuring density and then measuring mass with an appropriate scale).
For this samples, I am having an extremely hard time getting results to be precise in the 2-12 uL range that I need to work with with Tecan 50 uL filtered tips. These are all wet contact dispenses (not trying to free dispense).
For the sample plate, we have the plate on a Bioshake set to 4C. I have:
The speeds for aspiration and dispense = to the volume that is being delivered ( 2 uL/sec aspiration speed/dispense speed for 2 uL, 8 uL/sec aspiration speed/dispense speed for 8 uL etc)
1000 ms delays on aspirate and dispense.
LAG of 5 uL and TAG of 1 uL.
From this, Ive tried:
Adjusting the dispense deceleration and acceleration from the default 40K uL/sec down to 4000 uL/sec to see if that would help.
Change Delays from 1000 ms to 3000 ms without effect.
Prewet the tips 1X (1 sec delay on all aspirate and dispense steps) via microscript changes with the target transfer volume.
Prewet the tips 3X (1 sec delay on all aspirate and dispense steps) via microscript changes with the target transfer volume.
Removed the LAG and tried Reverse pipetting by overaspirating 1 uL.
Removed the LAG and tried Reverse pipetting by overaspirating 3 uL.
With all of these various approaches, I am unable to get a CV less than 5%. For example, at 8 uL, Ive had results that differ by 1 uL or more between replicate readings. None of the approaches above seem to yield better results than the others. All of the results look clean too (no droplets seen on the tips before/after aspiration/dispense. All 8 channels are used to aspirate and dispense, and the average volume of that well is calculated from the mass.
Am i missing some obvious strategy or liquid class setting that would help make these wet contact dispense more precise? I know this is a challenging liquid type, but I cant think of any other approaches to take than what I have tried already (pipette slow, long delays, try reverse pipetting).
Fair question. In the Fluent manuals, for 50 uL tips, the listed CV for a multi free dispense of a MasterMix (50% glycerol) with 50 uL tips is 5% for 5 uL and 3 % for 10 uL. Since this the closest to what I am working with, and a single contact dispense should be better than a free dispense, I’ve been using this as my goal (2-9.99 uL at 5% CV, 10-12 uL at 3% CV).
For the various strategies I’ve attempted (using 10 uL on the higher end of this range for testing, although since the accuracy is not adjusted yet, it ends up running anywhere from 7.23-9.79 uL by mass depending on the strategy tested):
Speeds for aspiration and dispense = the volume that is being delivered ( 2 uL/sec aspiration speed/dispense speed for 2 uL, 8 uL/sec aspiration speed/dispense speed for 8 uL etc), 1000 ms delays on aspirate and dispense, LAG of 5 uL and TAG of 1 uL, 1X Prewet on the aspirate. % CV = 3.73 % . This is probably the best result in limited testing, but I’m not confident it will hold up over more replicates, and a total difference of .56 uL is not going to be good enough for lower volumes to meet CV specs (and makes it hard to adjust the accuracy curve on top of that).
These speeds and air gaps used for all subsequent testing (except for reverse pipetting type testing).
Prewet the tips 3X (1 sec delay on all aspirate and dispense steps) via microscript changes with the target transfer volume - % CV = 5.73 %
Return Prewet to 1X, removed the LAG and tried Reverse pipetting by overaspirating 1 uL. % CV = 5.73 %
1X Prewet, Reverse pipetting by overaspirating 1 ul, Adjusting the dispense deceleration and acceleration from the default 40K uL/sec down to 4000 uL/sec to see if that would help - CV of 4.98%
3X Prewet, Reverse pipetting by overaspirating 1 ul, Adjusting the dispense deceleration and acceleration from the default 40K uL/sec down to 4000 uL/sec to see if that would help - CV of 5.90%
X Prewet, Reverse pipetting by overaspirating 3 ul, Adjusting the dispense deceleration and acceleration from the default 40K uL/sec down to 4000 uL/sec to see if that would help - CV of 10.16%
For another interesting point, we had done some initial tests with precision on this plasma/glycerol solution where the Bioshake was mistakenly turned off (so liquid at RT). When testing at 6 uL, with pretty much the same settings as my first test (only difference was aspirate/dispense speed were a little higher at 12 uL/sec), the uncorrected results were closer to 4 uL, but the precision was much better at 1.58%.
This led us to proceed with main adjustaccuracy() function adjustments for the cure, but we got bogged down in corrections not making sense. This led us to check in on precision again, and when we checked, the CV was 14.35% for 10 uL (1 result of 6.87 uL really threw off the results there). Which led to all the testing approaches I took above.
The only thing i can think of is slowing down the delays even >1 second (thought I tested that already), but when the lab pipettes this manually, it doesn’t seem to take > 1 second to stabilize in the tip, so not sure if that will actually help.
Any advice is appreciated that will give me better precision so I can use the adjust accuracy function!
Would you mind sharing the tip type that you are using (size, bore) and arm (FCA, MCA)?
Also if you can share the LC xml in a message, I’m happy to review it when I have some downtime during lunch hour. In the past I’ve been mindful of items like backlash compensation or even the speed of retraction from the container.
One thing I will add is that at smaller volumes, I’m hard pressed to trust data coming directly from companies. For starters, my lab is different than their testing labs. I’ve seen things like elevation, proximity to large bodies of water, or even floor level impact systems. Don’t get me started on HVAC systems. Second, they are experts at their platforms and can reach out to the folks who built the components (if they did so in-house) or can exploit novel features.
At something like 2uL, I would aim for a much broader CV % or at least push for guardbanding studies of ranges if possible. It is true that our LC testing is generally more specific to the moment (place & time) of the conditions, things can or will evolve and may be different at 11am vs 2am in 24/7 labs. Knowing the impact of a range on results is useful for troubleshooting but also can help inform your true CV % for operational efficiency.
With that said, I found some useful threads on glycerol strategies.
I agree on your points overall; I’m going to have the people in lab repeat this experiment manually to measure by mass, and see what their CV comes out to be. That could help make the case for broader CV. Its an easier conversation to have if we could meet those specs first, so I was trying to see if I was missing something obvious to make this work a little better first before I push on that boundary.
On aspirate and dispense, I have it coming out the tip at 20 mm/s and have a submerge depth of 1 mm. When I look at the the tips visually after aspirate and dispense, I don’t see any on the outside of the tip that would have indicated transferring extra volume on the outside.
I haven’t touched the backlash compensation, but I’d love get some recommendations for how to adjust that. While I get the concept of it, the manual has very little guidance on recommended settings, and I don’t see backlash compensation discussed on here anywhere else. Have you seen any generic relationship between the target volume to deliver and what you set the backlash compensation volume to be? It’s currently set to 5 uL backlash air volume and 3 ul backlash for volumes<-5 uL, so there would have been no backlash for all these 10 uL tests i ran.
Thanks @luisvillaautomata for the shout - I’ll see what I can recall immediately and check my notes next week.
We have a liqFCA with the metal 1mL tips, and in testing my solution (50% glycerol, 50% aqueous) we had to keep it chilled while performing the run. The really slow speeds and introducing a wash step every so often (once in the beginning, once after any mix, and another at the end for good measure) got us pretty close. It resets the system trailing air gap and that proved to help me a great deal.
What @HarenArul mentions in this comment on my post from back in 2023 about overdispensing intentionally, only to aspirate off any residual droplets formed is interesting and would require some additional testing, but it definitely seems like a solid play overall.
IIRC my lowest volume was also 2 uL, and I don’t believe we needed to hit below 5% CV or inaccuracy.
Thanks for the mention @Kastronaut. It does take a fair amount of testing but this can be shorten by building in a liquid detection data export to support optimization and only applies to dry dispensing applications.