My employer bought a MagNA Pure 96 instrument at auction. It doesn’t look like we’re going to have the nucleic acid extraction throughput to justify refurbishing it and using it for its intended purpose. But it also looks like it can do lots of things that would be useful as a general-purpose liquid handler.
Roche obviously does not want us to do this. All of the available protocols are kit-specific and there’s no method editor. It’s all meant to be very GxP and Audit Trail.
I’m not especially hopeful that this would be worth my time (read: my employer’s money), but if I wanted to build a direct control interface for this thing, where would I even start? People have obviously built their own interfaces for other liquid handlers - does that usually involve the support of the instrument’s manufacturer?
Now that I’m typing this out the safety issues alone feel like enough of a deterrent. But I’d still be interested in people’s thoughts on the issue.
Run a few kit protocols and capture the packet traces with Wireshark or a similar tool. Have claude code correlate firmware sent over the wire to the logged command being performed. Run some very basic and safe unit tests to make sure you’re directionally correct on command formatting. This is basically how pylabrobot was made but what took 6 months in 2022 takes 6 hours in 2026. Claude code is really really good at this but I would still take your time running a lot of tests manually and building solid infrastructure for distinguishing what is verified vs what is not.
What’s you’re break-even point on this project? Like, how much time would you have to spend converting this into a generic liquid handler before its not worth it? Also I dont understand why if it’s not suitable for nucleic acid extraction it would be suitable for other workflows.
The main cost factors are that the instrument knows it’s overdue for service and also refuses to run without scanning barcodes on unexpired reagents that we don’t have. Paying for those would already represent a decent fraction of the cost of a refurbished STAR. But it does move during initialization, so maybe that’s enough for a start.
I’d imagine the thing would work for nucleic acid extraction after servicing it and getting the right reagents, we just have a cheaper and less potentially broken way to do that. I want an automated pipetting tool that can combine excipients in gradients for formulation screening, pretty basic stuff as lab automation goes.
They actually lock you out if you’re overdue for service? Which I’m assuming you have to pay them to perform? That’s another level. I guess that would make it hard to even run the kit protocols to get the packet traces but there has to be some register value or cached data somewhere that you can modify to let you run your machine, basically trick it into thinking the service was performed more recently. The people that wrote these programs were not NSA cryptographers - most likely the data you’d have to change is somewhere accessible.
This is due to IVDR or similar regulations that the Magnapure is meant to adhere to. There are also IVD Hamilton STARs available, that “feature” a protocol abortion if you dare to pull a carrier out during the run. A pain in the behind, but solvable with new firmware, which is only available to Hamilton tech people, I think. Years ago I was able to break free of those restrictions by purchasing a software update with included firmware update.
If Hamilton was willing to do this on a Magnapure, that would be the route I would explore had I purchased such an instrument.
Hamilton’s products were used widely in Roche Diagnostic (especial for molecular diagnostic). You can find that many products (Hamilton STAR in cobas s 201, cobas x480, cobas p680, cobas p630 and etc) are all Hamilton STAR or STARlet.
