AI supports the discovery of highly binding antibodies

Scientists at the University of California San Diego School of Medicine have developed an artificial intelligence (AI) strategy for discovering high-affinity antibody drugs.

In the study, published Nature Communication, researchers used this technique to identify a new antibody that binds to a major cancer target 17 times faster than an existing drug. The authors say the pipeline could accelerate the discovery of new drugs against cancer and other diseases such as COVID-19 and rheumatoid arthritis.

To be a successful drug, an antibody must bind tightly to its target. To find such antibodies, researchers often start with a known amino acid sequence and use bacterial or yeast cells to produce a series of new antibodies with variations of that sequence. These mutations are tested for their ability to bind the target antigen. The small group of antibodies that work best is put through another round of modifications and tests, and this cycle is repeated until a group of hard-binding finalists emerges.

Despite this long and expensive process, many of the antibodies produced still fail to work in clinical trials. In a new study, UC San Diego scientists have developed a state-of-the-art machine learning algorithm to accelerate and streamline these efforts.

The process begins in the same way, with researchers generating an initial library of about half a million antibody sequences and screening them for their affinity for a specific protein target. But instead of repeating this process over and over again, they feed the dataset to a Bayesian neural network that can analyze the data and use it to predict the binding relationship of other sequences.

“With our machine learning tools, these next steps of mutation and sequence selection can be done quickly and efficiently on a computer rather than in a lab,” said senior author Wei Wang, PhD, professor of Cellular and Molecular Medicine at UC San Diego. School of Medicine.

Another advantage of their AI model is its ability to report the veracity of each sentence. “Unlike many AI methods, our model can tell us how confident it is in each of the predictions, which helps us place the antibodies and decide which ones we should prioritize in drug development,” Wang said.

To validate the pipeline, project scientists and first study authors Jonathan Parkinson, PhD, and Ryan Hard, PhD, set out to design anti-anti-death ligand 1 (PD-L1), a protein that shows It’s a lot of cancer. the target of many commercially available anticancer drugs. Using this method, they identified a new antibody that bound PD-L1 17 times better than atezolizumab (brand name Tecentriq), a wild-type antibody approved for clinical use by the Administration of the US Food and Drug Administration.

Researchers are currently using this method to identify promising antibodies against other antigens, such as SARS-CoV-2. They are also developing AI models that analyze the amino acid sequence for other antibody properties that are important for the success of clinical trials, such as stability, solubility and selectivity.

“By integrating these AI tools, scientists may be able to perform an increasing proportion of their antibody discovery efforts on a computer rather than at the bench, which could lead to of faster and smaller detection,” said Wang. “There are so many applications for this pipeline, and these studies are just the beginning.”

Reference: Parkinson J, Hard R, Wang W. The RESP AI model accelerates the identification of strongly binding antibodies. Nat Comms. 2023;14(1):454. two: 10.1038/s41467-023-36028-8.

This article has been reprinted from the following sources. Note: Information may be edited for length and content. For more information, please contact the source mentioned.