Machine learning uncovers potent immunomodulators for vaccines and immunotherapy

Small molecules called immunomodulators can assist create simpler vaccines and stronger immunotherapies to treat cancer.

But finding the molecules that instigate the correct immune response is difficult -;the variety of drug-like small molecules has been estimated to be 10⁶⁰, much higher than the variety of stars within the visible universe.

In a possible first for the sector of vaccine design, machine learning guided the invention of recent immune pathway-enhancing molecules and located one particular small molecule that would outperform one of the best immunomodulators available on the market. The outcomes are published within the journal Chemical Science.

“We used artificial intelligence methods to guide a search of an enormous chemical space,” said Prof. Aaron Esser-Kahn, co-author of the paper who led the experiments. “In doing so, we found molecules with record-level performance that no human would have suggested we try. We’re excited to share the blueprint for this process.”

“Machine learning is used heavily in drug design, but it surely doesn’t appear to have been previously utilized in this fashion for immunomodulator discovery,” said Prof. Andrew Ferguson, who led the machine learning. “It’s a pleasant example of transferring tools from one field to a different.”

Machine learning to screen molecules

Immunomodulators work by changing the signaling activity of innate immune pathways throughout the body. Particularly, the NF-κB pathway plays a job in inflammation and immune activation, while the IRF pathway is important in antiviral response.

Earlier this yr, the PME team conducted a high-throughput screen that checked out 40,000 mixtures of molecules to see if any affected these pathways. They then tested the highest candidates, finding that when those molecules were added to adjuvants -; ingredients that help boost the immune response in vaccines -; the molecules increased antibody response and reduced inflammation.

To seek out more candidates, the team used these results combined with a library of nearly 140,000 commercially available small molecules to guide an iterative computational and experimental process.

Graduate student Yifeng (Oliver) Tang used a machine learning technique called energetic learning, which blends each exploration and exploitation to efficiently navigate the experimental screening through molecular space. This approach learns from the information previously collected and finds potential high-performing molecules to be tested experimentally while also stating areas which were under-explored and should contain some useful candidates.

The method was iterative; the model identified potential good candidates or areas during which it needed more information, and the team conducted a high-throughput evaluation of those molecules after which fed the information back into the energetic learning algorithm.

Molecules that outperform the remainder

After 4 cycles -;and ultimately sampling only about 2% of the library -; the team found high-performing small molecules that had never been found before. These top-performing candidates improved NF-κB activity 110%, elevated IRF activity by 83%, and suppressed NF-κB activity by 128%.

One molecule induced a three-fold enhancement of IFN-β production when delivered with what’s called a STING (stimulator of interferon genes) agonist. STING agonists promote stronger immune responses inside tumors and are a promising treatment for cancer.

The challenge with STING has been that you may’t get enough immune activity within the tumor, or you may have off-target activity. The molecule we found outperformed one of the best published molecules by 20 percent.”

Prof. Aaron Esser-Kahn, co-author of the paper

Additionally they found several “generalists” -; immunomodulators able to modifying pathways when co-delivered with agonists, chemicals that activate cellular receptors to provide a biological response. These small molecules could ultimately be utilized in vaccines more broadly.

“These generalists might be good across all vaccines and due to this fact might be easier to bring to market,” Ferguson said. “That is quite exciting, that one molecule could play a multifaceted role.”

To raised understand the molecules found by machine learning, the team also identified common chemical features of the molecules that promoted desirable behaviors. “That enables us to concentrate on molecules which have these characteristics, or rationally engineer latest molecules with these chemical groups,” Ferguson said.

The team expects to proceed this process to go looking for more molecules and hope others in the sector will share datasets to make the search much more fruitful. They hope to screen molecules for more specific immune activity, like activating certain T-cells, or find a mix of molecules that provides them higher control of the immune response.

“Ultimately, we wish to search out molecules that may treat disease,” Esser-Kahn said.

A team from the Pritzker School of Molecular Engineering (PME) at The University of Chicago tackled the issue by utilizing machine learning to guide high-throughput experimental screening of this vast search space.