Exploring The Impact Of Automation And AI On Gene Therapy
By Life Science Connect Editorial Staff
The rapid advancement of automation tools is changing industries everywhere, including gene therapy research and development. In a recent episode of Cell & Gene Live, host Tyler Menichiello delved into best practices for aligning automation technologies with the short-term and long-term goals of gene therapy companies. Joining him were guest speakers Matt Hume, the head of lab automation at Beam Therapeutics, and the CSO of Excision BioTherapeutics, Dr. TJ Cradick.
Cutting-Edge Technologies
The speakers began by discussing the types of automation that have had the greatest impact on the industry in recent years. Humes praised the rise of collaborative robotics, which allow researchers and scientists to interact directly with the machines, rather than control them from behind protective barriers. He also noted how mobile robotics are helping connect islands of automation. “There are examples of mag-lev plate shuttles, high-speed linear rails,” he said. “At Beam we have these rovers that are essentially cars that use machine vision to navigate by bar code across our entire lab space, and they connect each of our systems via a handoff to the actual robotic arms.”
Cradick also lauded the range of innovations that “increase volume, speed results, and improve the quality and reliability of results.” But most impactful to him is the incorporation of artificial intelligence (AI), machine learning (ML), and automated computation. He admits that many of these are expensive for small and startup firms, but the sheer range of options means “many people can look at what they're doing and therefore decide what they can automate that will be the advancement that's most relevant to their work.”
Humes mentioned how his firm is using these technologies to apply large data sets for model training. “Beam is a base editing company, so some examples of applications in our area revolve around optimizing on-target edits,” he explained, such as “the design of the molecule, the sequence that we're using to guide our actual editor, and even optimizing different delivery modalities.”
While Excision is using AI and ML to improve experiment design and gene targeting protocols, what really impresses Cradick is how the technology “is now immersed into data automation coupling with our systems for doing these kinds of LIMS (laboratory information management system) things… that wasn't really thought about too long ago but is really changing things up.”
When To Start Thinking About Automation
“There’s no wrong time to start” automating, insisted Humes. At Beam, he said, the company has a lot of faith in automation, as it has since the beginning. “Our scientists had really ambitious plans,” he said. “I mean, still to this day, but especially in the beginning. And it just worked out great to have somebody there at those early conversations to say, ‘Hey, if you need to do this number of samples, we can set up this device for you and then, instead of making it a logistics question make it a how-can-we-do-the-best-science question?’”
Of course, Cradick noted that the right time to adopt automation hinges on the ability to afford it. That said, he emphasized the importance of calculating the potential ROI of a given solution. “Sometimes, just the miniaturization gives you a return on investment very quickly,” he explained. “If you use radiation expensive proteins or patented reagents, which is something that is applicable to our field, it really makes these calculations add up.”
Both speakers noted, beyond cost ROI, there are a number of “unintended benefits” to automation that drive early adoption, such as keeping employees engaged by eliminating repetitive tasks. Cradick added that there’s a certain appeal to seeing automation in action that attracts venture capitalists, business partners, or other visitors to the lab.
Navigating The Risks
Adopting automation is not without risks, the speakers admitted. Cradick mentioned that if standard operating procedures (SOPs) and protocols are clear, it can result in inconsistent or unreproducible results. Humes agreed, stating “Robots don’t fix bad science.” If the process isn’t reproducible or ready for high throughput, it’s just going to generate a lot of bad data.
Another risk to be aware of, according to Cradick, is relying too much on a single solution or unit for any given workflow lest it become a “doorstop.” Humes echoed this sentiment while stressing the importance of having redundancy to avoid creating a single point of failure.
Picking Technology Partners
The speakers offered advice for choosing automation technology partners. Cradick highlighted the importance of ensuring that the solution can do the tasks it needs to do. At the same time, he also emphasized prioritizing features. Recalling his own experience investigating a solution, he said, “there wasn't something that would do all of the above. Sometimes they couldn't grab the extra head, but they didn't have the ability to interface with the plates. So, then we said, ‘All right, what parts do we give up?’”
Humes said his interaction with vendors has changed over the years. “In my very first system, I was 100 percent reliant on the vendor and I kind of sat back, tried to learn as much as possible, and crossed my fingers most of the time hoping it all worked out,” he said. “Nowadays, with my experience and the experience of my team, we essentially send them a list of stuff that we want, where we want it, and then ask, ‘How fast can you get it to us?”
He added, “Having a partner that's willing to work on any end of that spectrum is really not only important but also telling in terms of their confidence of their own ability and their own product.”
When it comes to scaling up, Cradick also suggested outsourcing until the process can be brought back in-house. “It's just one more avenue that, particularly early or when you're doing some steps that are a scale or more above what you're able to do… figure out who are automation experts that you can outsource to.”