San Diego: a method to find high affinity antibody therapies which is based on artificial intelligence (AI) has been created by researchers at the University of California San Diego School of Medicine.
In the study findings published in Nature Communications, the researchers used the approach to identify a new antibody that binds to a major cancer target 17 times stronger than an existing one. antibody drug. The authors say the pipeline could speed up the discovery of new drugs against cancer and other diseases like COVID-19 and rheumatoid arthritis.
To be a successful drug, an antibody has to bind strongly to its target. To find such antibodies, researchers often start with a known amino acid sequence of the antibody and use bacterial or yeast cells to produce a series of new antibodies with variations of that sequence. These mutants are then evaluated for their ability to bind the target antigen. The best-performing subset of antibodies is then put through another round of mutations and evaluations, and this cycle is repeated until a set of closely related finalists emerges.
Despite this lengthy and expensive process, many of the resulting antibodies are still not effective in clinical trials. In the new study, UC San Diego scientists designed a next-generation machine learning algorithm to accelerate and optimize these efforts.
The approach begins similarly, with researchers generating an initial library of approximately half a million potential antibody sequences and testing them for their affinity for a specific target protein. But instead of repeating this process over and over again, they feed the data set into a Bayesian neural network that can analyze the information and use it to predict the binding affinity of other sequences.
“With our machine learning tools, these subsequent rounds of mutation and sequence selection can be carried out quickly and efficiently on a computer rather than in the lab,” said lead author Wei Wang, PhD, Professor of Medicine. Cellular and Molecular at UC San Diego. Medicine School.
A particular benefit of their AI model is its ability to report the accuracy of each prediction. “Unlike many AI methods, our model can actually tell us how confident it is in each of its predictions, which helps us rank antibodies and decide which ones to prioritize for drug development,” Wang said.
To validate the pipeline, project scientists and study co-authors Jonathan Parkinson, PhD, and Ryan Hard, PhD, set out to engineer an antibody against programmed death ligand 1 (PD-L1), a protein highly expressed in cancer. and the target of several commercially available anticancer drugs. Using this approach, they identified a new antibody that bound PD-L1 17 times better than atezolizumab (brand name Tecentriq), the wild-type antibody approved for clinical use by the US Food and Drug Administration.
Researchers are now using this approach to identify promising antibodies against other antigens, such as SARS-CoV-2. They are also developing additional AI models that analyze amino acid sequences for other antibody properties important for clinical trial success, such as stability, solubility, and selectivity.
“By combining these artificial intelligence tools, scientists can perform an increasing portion of their antibody discovery efforts on a computer rather than at the bank, which could lead to a faster discovery process that is less prone to errors.” failures,” Wang said. “There are so many applications for this pipeline, and these findings are really just the beginning.”
