Hansa Biopharma, the Swedish company behind the immunoglobulin G (IgG)-cleaving enzyme imlifidase, will use the AI protein-engineering platform from Dutch-Swiss firm Cradle to design its next generation of autoimmune drug candidates, the companies announced Oct. 1.
Hansa’s scientists will combine Cradle’s tools with Hansa’s existing experimental data, disease knowledge and protein-engineering expertise.
“We’re not building any of these AI modeling systems on our own,” said Sofia Järnum, Hansa’s vice president and head of research and early development. “So we really want to access the top AI technology by partnering.”
Hansa selected Cradle partly for its ability to optimize several protein properties at once, Järnum said. In protein engineering, each design cycle means making candidate molecules, testing them in the lab and feeding the results into the next round of designs. Balancing properties such as activity, stability and manufacturability can stretch that process across many rounds.
Sofia Järnum, Ph.D.
Cradle provides scientific advisers who help customers plan experiments and interpret results, said Stef van Grieken, Cradle’s CEO and co-founder. The stakes for each round can be substantial. He estimated that a three-month round of lab work can consume hundreds of thousands of dollars in reagents and materials.
How many months can AI save?
Cradle, which signed a three-year collaboration with Bayer in January, works with 10 of the top 25 global pharma companies, van Grieken said. Clients use its platform for applications ranging from initial hit identification to late lead optimization. In the announcement, Cradle said customers report shortening development timelines by up to 18 months, though how much time the platform saves depends on the use case.
“We work on anything from therapeutic enzymes to antibodies to CRISPR systems to cell therapies, and these by themselves carry different timelines,” van Grieken said. “If we take antibodies, what we’re seeing is that people go roughly twice as fast as they did before. So if an antibody program took 18 to 24 months, it’s now more like 12. It depends a little on what you’re doing, but 2x is pretty consistent.”
The 18-month saving cited in the announcement came from a vaccine project, van Grieken said.
“The strategy from the Hansa side is to use external technology around AI and complement it with our deep knowledge: the data we already have, the disease areas and our protein engineering expertise,” Järnum said.
Different projects may call for different tools, she added, and Hansa plans to apply Cradle to specific challenges where multi-parameter optimization is the main draw. “We hope it will make it possible to speed up the iterative rounds, so that they won’t be as delayed as they otherwise would be,” she said.
Moving more screening to the computer screen
Stef van Grieken
Van Grieken said the platform changes how teams plan a design round. Before committing to a set of candidates for lab testing, a team can weigh trade-offs among activity, melting temperature and potential liabilities. “You can still debate in silico with your team: should we go with this set, that set, or the other?” he said. Some customers now bring in therapeutic-area colleagues to weigh in on those trade-offs, he added. “So the process of getting to a molecule becomes a little more iterative and a little more like how software works. Not there yet, obviously, but there’s a lot more debate ahead of time.”
Van Grieken described a traditional route in which teams start from a natural sequence or immunize an animal, pick hits and screen down to a preclinical candidate. Earlier uses of AI in protein engineering, such as structure-prediction tools like AlphaFold or models that score sequences for properties like thermal stability, still left the scientist to decide which edits to make. “They’re the generator, if you will,” van Grieken said. Cradle’s platform generates the candidate sequences, balancing the properties a team specifies. “You might not have a screening team per se anymore, or a lead optimization team,” he said. “You might have people who move with their molecule through the whole process.”
Why the wet lab still holds the keys
The people who run the Cradle platform are mostly scientists who understand the molecules and disease mechanisms, van Grieken said. He contrasted this with competitors whose tools, by his account, can only be operated by ML experts or bioinformaticians. Computational teams sometimes connect through Cradle’s API instead. “We think they’re the best equipped to make decisions on the molecule,” he said of scientists.
Järnum said Hansa scientists will run the Cradle platform, while colleagues from chemistry, manufacturing and controls (CMC), translational science and the clinical teams follow the progress.
“It’s still the wet lab that is the bottleneck in this, as far as I can tell, and it probably will be for a while longer,” Järnum said.
Meanwhile, imlifidase, Hansa’s lead product, is commercially available in Europe as Idefirix and is under FDA review for kidney transplant desensitization, with a decision expected Dec. 19.
Tell Us What You Think!
You must be logged in to post a comment.