University of California San Diego School of Medicine researchers have devised an artificial intelligence (AI)-based technique for identifying high-affinity antibody therapeutics.
The method was utilised in the study, which was published on January 28, 2023 in Nature Communications, to develop a novel antibody that binds a significant cancer target 17-fold tighter than an existing antibody therapy. According to the scientists, the pipeline might hasten the development of new treatments for cancer and other disorders such as COVID-19 and rheumatoid arthritis.
An antibody must attach strongly to its target in order to be an effective medication. Researchers often begin with a known antibody amino acid sequence and utilise bacterial or yeast cells to create a series of novel antibodies with variations on that sequence. The ability of these mutants to bind the target antigen is subsequently tested. The best-performing antibodies are then submitted to another round of mutations and assessments, and the process is repeated until a group of tightly-binding finalists emerges.
Despite this lengthy and costly procedure, many of the resultant antibodies are ineffective in clinical trials. In the current study, UC San Diego researchers created a cutting-edge machine learning system to help expedite and simplify these activities.
Similarly, researchers generate an initial library of around half a million potential antibody sequences and test them for affinity to a given protein target. Instead of repeating the procedure, they input the dataset into a Bayesian neural network, which analyses the data and uses it to predict the binding affinity of additional sequences.
“These successive rounds of sequence mutation and selection may be carried out rapidly and effectively on a computer rather than in the lab,” said lead author Wei Wang, Ph.D., professor of Cellular and Molecular Medicine at the University of California, San Diego School of Medicine.
The capacity of their AI model to disclose the confidence of each forecast is a distinct benefit. “Unlike many AI systems, our model can tell us how confident it is in each of its predictions, which helps us rank the antibodies and select which ones to prioritise in drug development,” Wang said.
Jonathan Parkinson, Ph.D., and Ryan Hard, Ph.D., project scientists and co-first authors of the work, set out to build an antibody against programmed death ligand 1 (PD-L1), a protein widely expressed in cancer and the target of numerous commercially available anti-cancer therapies. Using this method, they discovered a new antibody that bound to PD-L1 17 times better than atezolizumab (Tecentriq), the wild-type antibody licenced for clinical use by the US Food and Drug Administration.
This method is currently being used by the researchers to find potential antibodies against additional antigens, such as SARS-CoV-2. They are also creating new AI models to examine amino acid sequences for other antibody qualities critical for clinical trial success, such as stability, solubility, and selectivity.
“By integrating these AI techniques, scientists may be able to undertake an increasing percentage of their antibody discovery efforts on a computer rather than at the bench,” Wang added. “This process has so many potential, and these discoveries are truly only the beginning.”
You might also be interested in reading, A ChatGPT bot “for professional usage” is on its way
