Robots and machine learning researchers combine forces to speed up the drug development process

IBM Research and Arctoris are bringing the power of artificial intelligence and robotic automation to the process of developing new drugs. The two companies aim to make smarter choices early on in the process, iterate faster and improve the odds of finding an effective treatment. 

IBM Research contributed two platforms to the project. RXN for Chemistry uses natural language processing to automate synthetic chemistry and artificial intelligence to make predictions about which compound has the highest chance of success. That information is passed on to RoboRXN, an automated platform for molecule synthesis.  

Arctoris, a drug discovery company, brought Ulysses to the project. The company’s automated platform uses robots and digital data capture to conduct lab experiments in cell and molecular biology and biochemistry and biophysics. Experiments conducted with Ulysses generate 100 times more data points per assay compared to industry-standard manual methods, according to Arctoris.

IBM Research will design and synthesize new chemical matter that Arctoris will test and analyze. The resulting data will inform the next iteration of the experiment.

SEE: Drug discovery company works with ethnobotanists and data scientists

Thomas A. Fleming, Arctoris co-founder and COO, described this project as “a world-first closed-loop drug discovery project” that combines AI and robotics-powered drug discovery.

“This collaboration will showcase how the combination of our unique technology platforms will lead to accelerated research based on better data enabling better decisions,” he said in a press release.

A research paper about closed-loop drug discovery describes the process as a centralized workflow controlled by machine learning. The system generates a hypothesis, synthesizes a lead drug candidate, tests it and then stores the data. This comprehensive process could “reduce bottlenecks and standards discrepancies and eliminate human biases in hypothesis generation,” according to the paper. 

Automating lab work results in better data which in turn means less rework and a savings of time and money, Poppy Roworth, head of laboratory at Arctoris, explained in a blog post. She described the benefits of automation this way: “I no longer have to manually pipette each well at a time of a 96 or 384 well plate, which is highly beneficial for my sanity when there is a stack of more than 5 or 10 to get through.” By automating the protocol, scientists can use time previously spent in the lab on “planning the next experiment, designing new projects with clients, reading literature and keeping up to day with other projects.” 

Matteo Manica, a research scientist at IBM Research Europe, Zurich, is coordinating the project and said in a press release that this work is a unique opportunity to quantify the impact of AI and automation technologies in accelerating scientific discovery. 

“In our collaboration, we demonstrate a pipeline to perform iterative design cycles where generative models suggest candidates that are synthesized with RoboRXN and screened with Ulysses,” he said. “The data produced by Ulysses will then be used to establish a feedback loop to retrain the generative AI and improve the proposed leads in a completely data-driven fashion.”

More than 3,000 researchers in 16 locations on five continents work for IBM Research. Arctoris is a biotech company headquartered in Oxford with offices in Boston and Singapore. The collaboration is ongoing.

Data, Analytics and AI Newsletter

Learn the latest news and best practices about data science, big data analytics, and artificial intelligence.
Delivered Mondays

Sign up today

Also see