AI-generated automated protocols for life sciences
“Without wanting to be hyperbolic, LLMs will revolutionize how we design automated protocols in the life sciences,” Brennan-Badal explained. Opentrons’ compatibility with LLMs makes it possible to create “AI-generated automated protocols for every well-documented manual experiment in the scientific literature, as well as every commercially available molecular biology kit,” Brennan-Badal added. This capability, he believes, will greatly influence the pace and magnitude of biological experiments.
Currently, the development and refinement of a single nucleic acid extraction protocol can take a significant amount of time, ranging from weeks to months, as researchers strive to optimize variables such as sample type, DNA yield and quality and reproducibility. “That process could be vastly improved using AI,” Brennan-Badal said.
Flex designed to support affordable and efficient lab automation
In addition to its AI focus, Opentrons strives to carve out a nice for itself in a crowded lab automation space with notable players such as Labcyte, Tecan and Thermo Fisher Scientific by emphasizing affordability and efficiency. “Regarding the overall expenditure, the Flex’s cost is about a tenth of what its market counterparts demand,” according to Brennan-Badal. He credits this cost reduction to the company’s use of vertical integration in manufacturing processes and supply chain management.
The Flex robot is designed to be set up within a day, Brennan-Badal said. The system can be reconfigured by swapping out pipettes and hardware modules in a short amount of time. He attributes this modularity to the product design, on-device display instructions and smart sensors supporting auto-calibration.
The company also designed the Flex to have the capacity to learn. Brennan-Badal explains that the robot’s open application programming interface (API) is compatible with “design-of-experiment” programs, including AI-driven agents to support future developments in AI integration.
“For software, we have a no-code protocol designer GUI, an open repository of verified and community-submitted protocols available for download, as well as a well-documented Python API for those that want ultimate control over their protocol, or to integrate third-party devices or software with the robot,” Brennan-Badal said.
AI in lab automation: Enhancing flexibility and experiment design
One of the factors stoking interest in generative AI platforms are their ability to code in a variety of languages. Flex takes advantage of this capacity, Brennan-Badal noted. “We have demonstrated that LLMs including ChatGPT 3.5 and 4 can write, and correct, executable Python code to automate experiments such as serial dilutions and aliquoting on Opentrons robots, and we’re beginning to get good data for more complex protocols,” he said.
Researchers at Carnegie Mellon led by Gabe Gomes created an AI agent that combined LLMs to design chemical synthesis experiments, and then used these experimental designs as input to generate Python code to successfully automate the reactions on an Opentrons robot,” Brennan-Badal noted. That research was highlighted in the preprint “Emergent autonomous scientific research capabilities of large language models.”
Opentrons’ open and well-documented API is compatible with such programs. “LLMs can be trained on the large corpus of Opentrons’ automation and protocol data to generate new automated experimental protocols,” Brennan-Badal asserted. He noted that this capability makes its platforms future-facing in terms of their ability to work within AI frameworks.
An open source philosophy
Open-source technology is a core element of the Flex robot’s design, which could have implications for the scientific community. According to Brennan-Badal, the open-source approach can support scientists by enabling platform development and integration, and by fostering the sharing of protocols and knowledge.
The Flex robot’s open-source ethos offers significant implications for the scientific community. Brennan-Badal opined, “Being open source supports scientists in two fundamental ways. It liberates scientists to do new things, including building on the platform and integrating with other hardware and software. And it facilitates sharing of protocols and knowledge, allowing scientists around the world to reproduce each other’s work.”
Brennan-Badal also highlighted the role Opentrons played during the early days of the pandemic, noting how its automation capacity streamlined laboratory workflows. He highlighted the use of the OT-2 model in Opentrons’ own Pandemic Response Lab in New York City, where it contributed to COVID sample processing.
Future of life sciences with AI and Lab automation
Looking ahead, the integration of artificial intelligence, specifically large language models (LLMs), with lab automation could have implications for the future of life sciences. Brennan-Badal suggests that the combination of Opentrons’ compatibility with LLMs and the development of AI-generated automated protocols may influence the way biological experiments are conducted in terms of speed and scale.
Filed Under: Cell & gene therapy, Data science, Industry 4.0, machine learning and AI, Omics/sequencing