How should we bring automation into the life sciences? Where should we start, and how can we get the most out of it? Synthace’s Life Science Business Director, Fane Mensah spoke with Bob Gantzer, previously Director of Lab Automation at Beam Therapeutics (at the time of the interview) and now VP and head of automation and lab technology at Aera Therapeutics, about the place of automation in biology.
It’s easy to make the mistake that adding automation into the life science toolbox comes as an “off the shelf” option. Worst case it might mean building a robot, or writing some code, right? But is it really just a matter of engineering? Well, no. Scientists are very much a part of automation. In fact, they are possibly the most important aspect.
Bob Gantzer’s early exposure to automation began with auto-pipetting in his early career as an attempt to make his life easier. It soon became clear that automation is not a panacea to the trials and tribulations of doing science.
“It's not as simple as just programming,” Gantzer says, “then a robot moves and all of a sudden ‘bam!’ science happens.” Automating science still requires having a basic understanding of the science involved. Automating the activities of science is essentially just adding another skill into the mix, such as learning how to program in Python.
Rather than repeating Gantzer’s experience of almost stumbling into the automation world, where should a life scientist begin?
Automation strategy: you can’t just throw people at the problem
Gantzer believes that many companies might think it’s easier to “throw people at the problem” in the short term. But building automation in as early as possible might prove more effective. “At Beam, the leadership made it a priority to have automation really embedded into our culture and a part of that was to bring me in as employee 26 or 27 to lead the automation team,” said Gantzer, “But I think what people are realizing is that once you get to scale it's hard to go back and automate it, there's just so much work to be undone at that point.”
Having a strategy is a vital piece of the automation puzzle. Integrating systems was part of Gantzer’s mission from early on.
“Looking at just how complex some systems are, integrated builds can take a year to realize,” he continued. “I mean you're looking at nine months to a year to integrate. I couldn't be hired at Beam and not deliver anything for a year.” That meant doing something, anything, but ultimately putting the pieces together. He started with one piece of equipment at a time. That meant getting it dusted off, on the bench, plugged in, and working.
That might mean not using the full capability of the system or its components to begin with. Instead, this approach grows over time with new ‘features’ being added as new components get included. With this approach, when it’s time to deploy the system, it’s already up and running, making the live process so much faster.
Automation buy-in... or automation bust
The biggest challenge, in Gantzer’s view, is getting scientists to buy in to using automation in the first place. “It’s a people challenge, not a technical challenge. The number one thing that I have to think about was to have this underscored by my boss on day one.” The challenge is effectively selling automation to your leadership.
After the selling comes the need to build momentum and experience. That’s when it becomes increasingly possible to show what the instruments can do. That confidence and experience facilitate different departments and groups working together to truly integrate.
When starting from scratch, the best approach requires buy-in from senior stakeholders but adding the ‘bottom-up’ angle maximizes the chances of successful integration. Gantzer added, “One of the things for us has just been to make the carrot for your scientists as big as possible.” Treating the scientists as customers is a useful analogy. They need to be sold on the features and benefits initially.
Better data has to be the goal
Gantzer paints an attractive scenario: “We can outline the vision: just imagine you can digitally design your experiment, you come in the lab you can load it, press go, and then you can go drink coffee for the afternoon, and then you come back to a beautiful dashboard of data.”
After all, data is paramount. Data is what scientists feed on. It’s what experiments generate and what allows us to understand our world. Automating the repetitive or time-consuming parts just makes that whole process better.
Data, of course, have to be presented in a usable format. Masses of data in a CSV file won’t inspire even the best of us. Effectively partnering scientists and automation engineers is the magic sauce that adds value to the output. Knowing how and where to add metadata to enrich the outputs makes that data so much more useable at the end of the day. It’s vital to know what the data means—even before it’s analyzed, and that requires some discussion up front—allowing scientists to give that all-important context to the automation engineers and show the scientists how the automation manages the experiments add the clarity that could so easily be missed by skipping this step.
This leads to the question of who to get involved. How do you build an effective team?
Building a team around automation
“None of us learn this in undergrad and grad school. It’s very, very rare to even have a standard liquid handler presented anywhere in training,” said Gantzer. “You almost have to accept that you're going to have an apprenticeship model to some degree, in order to build up the talent pipeline.” Currently, those coming into this side of the industry are still poorly represented on the engineering and software side, with around a third coming from these areas.
“For lab automation to grow effectively, that balance needs to shift. More formalized exposure to integrate liquid handling is one possible approach, enhancing the likelihood of embedding the necessary skills in scientists," as Gantzer adds, “Then, it’s just a matter of changing the mindset of the scientists doing the experimentation to embrace these approaches.”
Another of the key benefits of automation comes in the shape of throughput and complexity. It’s not necessary to only think in terms of high throughput to extract any value from automation. What works for thousands of samples, also applies to a few. Running all experiments from automated systems ensures that the data – all the data – remains within the system, is formatted consistently, and can be accessed or replicated by anyone at any time.
Can you really have it all?
Gone are the days of scientists running small-scale benchtop experiments but forgetting to upload experimental data to their LIMS system or worse, losing it. Instead, automation offers researchers the opportunity to explore more variables, in a shorter timeframe, and produce a richness of data not practically achievable by traditional, manual approaches.
But realizing automation’s full potential requires some planning. And it’s not only the technical and engineering aspects that need to be considered. People are an integral and crucial part of that planning. Winning the hearts and minds of senior stakeholders and bench scientists can, sometimes, be straightforward. The benefits of automation do the sales pitch themselves.
Want to listen to the interview this article is based on? Make sure to check out the Synthace podcast.
