Since raising a mind-boggling $1 billion in 2024, AI-driven Xaira Therapeutics has remained largely silent on where that money is being put to work. Last week, the company’s Chief Operating Officer, Jeff Jonker, opened up to Fierce Biotech, divulging details about the secretive biotech’s efforts in inflammatory and immunological (I&I) science.
“We’ve been a little tight-lipped, I’ll acknowledge, but I think that’s all about to change,” said Jonker, who joined the company last summer.
Xaira is the seventh venture-backed startup Jonker has worked for since leaving Genentech in 2009. During his time at the now Roche-owned subsidiary, Jonker worked with Marc Tessier-Lavigne, Ph.D., who currently serves as Xaira’s CEO.
“We are actively working on building a pipeline,” Jonker continued. “Part of what makes us a next-gen biopharma company is that the AI platform came first and then the pipeline that it generates will come second.”
Xaira’s roots come from co-founder and Nobel Prize winner David Baker, Ph.D., a biochemist and professor with the Institute for Protein Design at the University of Washington School of Medicine. His work includes designing models that predict or design the types of antibodies that bind to proteins.
“That’s where we started, so we are working on building antibody therapeutics,” Jonker explained. “What we’re doing is kind of developing the technology to work on targets that everybody agrees would be really compelling to test in humans, but nobody’s been able to crack.”
Related
“The idea is this: the system that runs us is so complicated and has so many variable points that are operating in parallel that it’s just beyond our human brains to be able to hold that many data points,” Jonker said.
The Xaira exec detailed how most of the data available about how cells function is observational, giving researchers the ability to describe basic functions, but failing to provide much insight regarding causal biology. In Jonker’s words, a lot of modeling data is “just capturing ‘what is’ and not the ‘why.’”
“What happens when this thing goes wrong? Or what is the way that you make something that’s going wrong go back to the healthy state?” he asked. Those are some of the questions Xaira is looking to answer.
One way they’re doing that is by “perturbing all the genes in a cell.” The company is then using that data to build and train machine learning models.
On Tuesday, Xaira unveiled an AI-powered cell model that the company called the largest of its kind to date. Known as X-Cell, the model is meant to predict the function of genes based on training data of genetic perturbations, where a gene is silenced and the resulting effects on the cell are catalogued.
“With X-Cell, we are already seeing the model make predictions that go beyond the data we provided,” Tessier-Lavigne said in a March 17 release. “That’s a big step toward a powerful model that will provide not just biological insights, but therapeutically actionable ones.”
Xaira plans to make some of the model and its underlying data available to other scientists, according to the release, and the biotech has published a non-peer-reviewed preprint (PDF) about X-Cell’s development.
Even with data from 25.6 million individual cells already in its databanks, Xaira plans to continue gathering more from different kinds of primary cells and stem cells, as well as from genetic perturbations of organoids and in vivo systems, the company said.
“It’s a very different kind of training,” Jonker told Fierce. “I don’t think it’s going to supplant scientists, but it is an augmentation to the way that we work that is unlocking really interesting connections.”
Related
“If you ask it, ‘Hey, you know, what do you think this gene does?’ in a cell type that it hasn’t seen, it’s actually beginning to have some sense of what the answer would be,” Jonker explained.
The tech is being trained on a variety of cells, including primary cell types from both healthy patients and those with certain diseases.
“The other things we can ask are like, ‘Hey, what are the things that activate T cells?’” Jonker continued. “There’s a number of canonical proteins that we know activate T cells. Those are playing a huge role in I&I diseases.”
When asked without being trained on the available literature, but solely on the cell data provided, the model can identify most known activators—plus a few unknowns.
“We’re still in the midst of experimentation on that, but it’s actually looking like it’s identifying proteins that are active in T cell activation that we didn’t know about,” he explained.
“That’s the beginning of what we’re hunting,” Jonker said. “The exciting thing is if we can begin to unlock pieces of biology that we just haven’t been able to unlock before, and then you can start doing it at a faster clip.”
Xaira is hardly alone in using AI to design or engineer antibodies. Flagship Pioneering’s Generate:Biomedicines, which recently went public, uses the tech to enhance aspects of its antibodies, as does Roche’s Japanese subsidiary Chugai Pharmaceutical.
And Chai Discovery, which also uses AI to brew up new antibodies, recently closed a $130 million series B to fuel its efforts.
“But as anybody knows who’s played in I&I, people are pretty heterogeneous,” Jonker said. Because of that, some drugs “work wonders for some patients,” while other patients with the same symptoms don’t respond.
“There’s obviously different biology underlying that,” the COO explained, adding that the industry just doesn’t know what it is.
“So, having an alternative set of targets that we didn’t know about—that would open up the opportunity to go and develop a medicine against those targets,” Jonker said. “That’s just the kind of opening that our industry is hungriest for.”
Related
“Our plan is to build a completely integrated R&D platform based on machine learning and then use that platform to generate new medicines that are profoundly different from the ones that are out there and that are otherwise unattainable without that technology,” Jonker overviewed, adding that Xaira will then want to test those in clinical trials.
“That’s going to take multiple years and it’s going to take a billion dollars—maybe more,” Jonker said.
The full interview with Jeff Jonker will air this Friday on The Top Line. Follow The Top Line on Apple Podcasts, Spotify, Amazon, iHeart Radio or wherever you get your podcasts.
