Bench
A typical individual lab bench has very little in the way of instrumentation. A set of pipettors, a vortexer, a minifuge, one or a few heating blocks - that's often it. At Warp I had a ceramic unit that enabled sterilizing steel inoculation loops.
Instrumentation at the bench scale - something that an individual can claim ownership - must be in very high demand and relatively low cost.
Walkup
When I was at Codon Devices, it came to pass that I was overseeing our attempts to remediate gene assembly attempts which had previously failed. This might entail sending it back for another attempt or just having more clones sequenced or sending an existing clone back through if one of the Sanger reads had failed. I did this by injecting new rows into a relational database, which would then trigger the work. That work would actually be executed by my lab co-workers following lists of instructions to load various instruments, run a specific protocol, and then unload that instrument and move the contents to a different instrument. So pull some plates, extract DNA, run Sanger reactions and load the Sanger instrument. Each of these stations was a walkup instrument.
Walkup instrumentation is most typically large and expensive, but specialized in a single task - a "unit operation". These might include centrifuges, reagent dispensers, liquid handling robots, thermocyclers - or any other instrument you might find in a lab. In many cases, a laboratory will host only one or a few of an instrument type. In other cases there might be more of an instrument than individual scientist benches - thermocyclers are useful to have in excess - but it still is valuable to have them as a shared bank rather than parceling them out to individual bench fiefdoms.
One constant bit of mental conflict for me with lab automation is three letters long. The standards for labware were established by a central body, which is how we have typical 96-well, 384-well and so forth plates. They could have been very different - when I was an intern nearly 40 years ago the immunoassays were run in plates that were akin to 4-6 modern plates joined on their long edge - huge boats. That would have made laboratory instrumentation even bigger. But anyways, the name of that standards body was the Society for Biomolecular Screening, so these plates are known as "SBS format" - but SBS is forever Sequencing by Synthesis!
Workcell
Back in my Codon example, samples were carried between the different instruments by human hands. But what if a robot did it instead? What if a set of instruments for a specific workflow were clustered in space so they could all be served by a robot arm? That is the nature of an integrated workcell.
Workcells are often used in high throughput screening, but can also be found supporting many other workflows - such as NGS or qPCR. Some are even set up for nearly continuous operation, if a human keeps stuffing input materials in one end and either the system can dump used materials or a human pulls out finished products.
Autonomous
Autonomous laboratories are the logical evolution of dedicated workcells. If instead of building a highly integrated system around a single workflow, what if one creates an orchestra of interconnected instruments which can support any workflow that uses any combination of unit operations enabled by those instruments? That's the idea of an autonomous laboratory. If you come to SLAS this year Ginkgo will be exhibiting a small system - I believe we're slated to have 10 individual stations (RACs = Reconfigurable Automation Carts). Alternatively, make an appointment with me to tour our headquarters and I think we're pushing 40 RACs. Those RACs include multiple types of liquid handlers (with different strengths and weaknesses), reagent dispensers, thermocyclers (standard and qPCR), plate reader, sealers and peelers, centrifuges, incubators, plate holders, and more. Pacific Northwest National Laboratories just ordered a system with over 90 RACs - who knows what the limit is?
Autonomous labs, at least with the right software and attention to engineering reliability, should be able to run at all hours with minimal human supervision (often just remote monitoring) - and run multiple independent protocols simultaneously. Truly autonomous labs could let klutzes like me run sophisticated molecular biology protocols like a pro, as well as enable "lab-in-a-loop" schemes in which an AI agent iterates through different versions of an experimental protocols, using the outputs from one run to assist in designing the next pass.
Blurring the Boundaries
As with any set of categories, there are cases which don't fit neatly into the scheme.
For example, many liquid handling robots for NGS applications are often tricked out with components to perform additional unit operations beyond pipetting. These often include thermocycling, magnetic bead operations and so forth, enabling full workflows from purified DNA to pooled library. So a single walkup instrument, but being used like an integrated workcell.
Similarly, any autonomous laboratory system can be used to design workcells, or an autonomous system might be configured to enable individual stations to be used as walkup instruments or portions might be set up more like integrated workcells.
Autonomous lab at Ginkgo HQ - though it has gone through more rounds of expansion since this photo was taken. Drop me a line if you'd like a tour!
1 comment:
This is a problem that has been solved for quite some time in the semiconductor industry. There is no need to reinvent the wheel. The main problem is that biotech volumes usually cannot justify the outlays required to bring a fully automated solution to market, nor can a customer usually justify an outlay of millions of dollars for lights out automation. That will remain the case until the industry moves away from the silly concept of a "shift" as a unit of work.
Post a Comment