I am a new user to both Hamilton Systems (Have used Opentrons a lot in the past) and pyLabRobot. For context I am also more of a molecular biologist then automation engineer, but have been tasked with protocol development for our STAR.
I wrote a normalization protocol that theoretically should variable volume repeat dispense diluent into an empty PCR plate then add the correct amount of sample into each plate. I was running water tests and was seeing variable volume levels at the end of the run. After further testing, I figured out that the first part (repeat dispensing diluent) was adding too much water to each well. When I fixed the concentrations to all be the same, it should have been adding 36uL but there was ~39-40 uL in each well.
I was not using a default hamilton liquid class, but I listed the settings I was using below:
flow rate: 100
aspiration settling time: 0.3
dispense settling time: 1
aspiration height: Using LLD
dispense height: fixed height of 1mm
Does anyone have any insight on if this is a liquid class issue or a hardware issue? The system was purchased used, but validated by a third party distributer. I am happy to share more details or the protocol itself if that would be helpful. Thank you all in advance!
One way to know if the instrument is in spec is to run the FVK (field verification kit). Hamilton uses scales to verify with a special liquid that your instrument is in spec.
I have no idea what those parameters you are using but typically I start with the default water then tune it gravimetrically with a scale. Elevation and humidity play a large role and liquid classes need to be tuned. Once tuned we fined in the same environment they transfer pretty dang well instrument to instrument.
Thank you that makes sense. After more digging I think the volume correction curve as causing me my issues. When I changed it from the default from the liquid class to a flat curve all my measurements were coming in more accurately. Some fine tuning is definitely still needed, but it is performing much better. I don’t currently have any scales that can connect with the hamilton, but I will look into it.
As background: no liquid handler is particularly accurate by default and we are building various PyLabRobot tools to enable more people to automate the generation of liquid transfer parameter sets (or in legacy term “liquid class”).
This is very important for accurate and consistent liquid handling. Especially if you are going straight for normalisation protocols (not easy) in which your automated Protocol nornally has to decide how much diluent to add into which well (in turn having to make independent decisions about tip choice, and liquid transfer parameters based on calculated diluent volume).
Depending on your setup you can start quickly or need to purchase equipment:
Gravimetric: buy Mettler Toledo WXS205 scale directly or Hamilton’s “Liquid Verification Kit” (see docs tutorial)
Absorbance: buy Tartrazine dye, use any absorbance plate reader if you have some standing around
Fluorimetry: buy any Qubit kit (DNA ones are easier to handle because they dont require -80 storage of reagents), use any fluorescence plate reader you might have standing around.
Each approach obviously requires more detailed information to actually fully automate and happy to discuss more but these are the starting points.
But if you need a specific volume only occasionally with standard liquid, we’re also happy to just let our autonomous agent run to create a new parameter set for you / cross-validate yours
Thank you for the advice. I didn’t realize the nuances of a normalization protocol before I started, but I am definitely getting there. Luckily for my applications ~10% CV is acceptable so I only need to get it “good enough”. I am going to attempt to build liquid transfer parameters using a scale we currently have in a more manual way, but if that doesn’t work I will definitely look into getting a Mettler Toledo WXS205.
I really appreciate the offer to cross validate! I am going to give it a shot with what I currently have, but if that doesn’t work I might take you up on your offer. Thanks again for the useful suggestions!
