Dr. Matthew P. Graving has been interested in using computer science in drug discovery for over 20 years. So much so that when Google first launched his email platform, he had an electronic address that fused computational science with biochemistry.
“I’ve been thinking for a long time about how to integrate biochemistry and computer science,” said Graving, vice president of platform technology and machine learning at iBio. Gen.
Mr. Graving, accompanied by CEO and CSO Martin Brenner, Dr. DVM, met with: Gen A temporary office near the back corner of a new open-concept start-up style office in San Diego, in an area with a coffee table and some chairs, and two long, shoulder-high bookshelves placed vertically surrounded by walls. Meetings are held outdoors, but not because of lack of space.
As a publicly traded company, iBio is innovating and staying as transparent as possible. “We think he probably started this turnaround story two years ago by hiring the right team and putting the right technology in place, and finally moved into new territory last year,” Brenner said. said. “We are in the process and final stages of switching gears completely to a new biotechnology,” he said.
In September 2022, iBio announced that it would acquire certain assets from RubrYc Therapeutics, a company co-founded by Graving and topping the artificial intelligence (AI) drug discovery platform and pipeline. Two months later, the company began the process of selling its CDMO business and cGMP biologics manufacturing facility, shifting investments to its immuno-oncology assets and AI-based drug discovery pipeline. The restructuring could result in cost savings of approximately 50% annually. A month later, CEO Thomas Isset stepped down as CEO and board member, and Brenner took over as interim CEO in addition to his role as iBio’s CSO.
Drug discovery through three actions
Both Graving and Brenner have backgrounds in mathematics and computer science, and are both calculated and cautious when discussing the use of AI. They argue that iBio uses machine learning to induce antibodies against specific protein epitopes, but what they really want to know is the real-world application. This technology stack also allows iBio to optimize these antibodies in a way that reduces downstream risk and speeds up the process.
They seem pretty reserved when it comes to potential innovations in machine learning that are built into nearly every aspect of their three-part discovery process.
First, they utilize an engine for engineering epitopes for elusive targets. About six years ago, Graving said of the computational tools he developed as the basis for RubrYc, “we’ve seen significant efficiency gains, especially for traditionally difficult goals.” He attributed the success to his chance discovery with Cody Moore, a young scientist who had risen at RubrYc and joined his Greving at iBio.
Second, iBio uses models based on natural language processing by leveraging several publicly available databases and proprietary in-house generated data on human protein structures. However, instead of dealing with the written word, it has been used for antibody sequences. Essentially, the processor spews out every computationally imaginable version of the antibody, allowing iBio to build an antibody library with full human sequence diversity.
“Our library may contain hundreds of thousands, or even millions, of antibody sequences,” Graving said. “He takes one or more engineered epitopes and incubates them with a library. These artificial epitopes can be made quickly and cheaply. can.”
To make a difficult problem even harder, Graving and Brenner are working on multispecific antibodies. “Maybe only 10 out of a billion people.” [antibody sequences] 10 in a billion is very hard to find using traditional methods. But if you really focus on it with an artificial epitope, you can extract the epitope more efficiently. “
Third, they developed mammalian display of these antibody libraries. This is important for scaling up antibody production. To achieve this, iBio’s pipeline integrates cells used in production to express antibodies early in the discovery process. Specifically, we engineer CHO cells to present IgG antibodies that mimic B cells and perform single-cell sorting to find cells that best express candidate antibodies.
“This accelerates what we can do downstream and allows us to quickly determine the feasibility of developing an antibody. will be gone,” he said. graving. “Literally within weeks, we can optimize and bring out antibodies that we think have high development potential.”
Mean and Lean Machine
iBio is currently in the preclinical stage and the most advanced program is in the IND realization stage. As part of our enhancements, we are leveraging external partners to help us focus on our team and pipeline.
“I’m a big fan of lean R&D,” says Brenner. “By focusing our team on what we are good at, we can actually run a relatively broad pipeline. and then run and optimize an in vivo screening model to see if there is an in vivo proof of concept before moving on to more complex translational animal models. We are able to make very rapid decisions to introduce the in IND readiness and process development.We are able to go from idea to in vivo proof of concept very quickly in less than six months.”
To cut down on waste, Brenner said he’s looking to hire “bilingual” scientists who are versed in both biology and machine learning or data science. “I think what makes us different and successful is that he basically integrated both sides of the equation into one person,” says Brenner. “For example, our director of machine learning, Alex Taguchi, doesn’t just design epitopes. He’s in the lab, too. He’s not just doing it, he’s also good at machine learning.” He said he tried to get a Gmail address and had to settle for a slightly different version.
Brenner said this kind of interdisciplinary scientist is essential for companies that use machine learning. “It’s very interesting to see first-generation AI-enabled companies still struggling,” Brenner said. “Big pharma is unlikely to be able to overcome the gap between these two areas. So if you build this as designed, you can avoid these silos. You can’t really explain the ambiguity of biological experiments, maybe five or six years ago there weren’t any such scientists, but there are now.”
This philosophy is literally embedded in the space of iBio, with half dry lab computations and half wet lab biochemistry and molecular biology, making each lab bench a step in the drug discovery process. Designed with care. to the next. The only separation is the floor-to-ceiling glass.
never surrender
Despite how humble Mr. Graving and Mr. Brenner are, they are equally ambitious and resolute. They are working harder and harder to prove what their platform can do. “So far we are not yet at the point where we have to raise the white flag. We are at least 8-8,” Brenner said. “That doesn’t mean we will continue to be 100% successful. But so far we have not reached a goal that the platform cannot solve.”
iBio may seem limited, but in reality, resources can be limited. Brenner and Graving would like to study infectious diseases and neurodegeneration, but the best application of their techniques is in immuno-oncology.
“By the end of this year, I want all antibody discoverers to know that if they have a hard problem to solve, they know where they’re going. There’s enough evidence in the pipeline. “It should be us, because there is,” Brenner said, “that we can really pursue very difficult drug targets that others have tried and failed.” .
Graving is concerned about the potential of machine learning in drug discovery. “My first job was during the internet boom and I saw internet companies with huge potential,” he said. “Well, that bubble almost burst. I don’t think we’re in this situation. [machine learning] bubble now. Thanks to machine learning, we are now able to discover new targets with high efficiency, but in the midst of it, we feel like we have to move! Become. “
