AI-Designed Antibodies: A Revolution in Antibody Drug Discovery

The use of artificial intelligence (AI) in antibody design has the potential to revolutionize the development of novel therapeutics. AI-based approaches enable the rapid identification of antibody sequences that bind to specific target proteins, offering a significant advancement over traditional methods.

One such method, RESP, developed by researchers at the University of California, San Diego (UCSD), uses a Bayesian neural network to predict the binding affinity of antibody sequences, accelerating the discovery of antibody drugs. The RESP pipeline identifies antibodies with high affinity and specificity, making it a valuable tool in cancer and rheumatoid arthritis drug discovery.

While traditional methods for antibody discovery involve immunizing animals or screening large numbers of molecules, AI-powered biotechs such as AbSci, AbCellera, Generate Biomedicine, BigHat Biosciences, and Biolojic Design are paving the way for faster and more cost-effective antibody discovery. These companies harness the power of AI to optimize antibodies for binding, solubility, yield, and immunogenicity.

DenovAI, a startup founded by former EMBL staff scientist Kashif Sadiq, aims to broaden the scope of antibody therapy by developing an AI-powered biophysics solution for antibody discovery. By combining advanced machine learning with computational biophysics, DenovAI hopes to significantly reduce the time and cost associated with identifying promising antibody candidates.

A significant breakthrough in AI-guided antibody design has been the development of RFdiffusion, a generative AI model based on the RoseTTAFold structure prediction network. Created by a team at the University of Washington, RFdiffusion excels in various design challenges, including unconditional protein monomer design, symmetric oligomer design, and enzyme active site scaffolding. The accuracy of this method has been validated through experimental characterisation of designed proteins, demonstrating its potential in drug discovery.

RFdiffusion represents a powerful tool for designing novel antibodies. In a study, researchers used RFdiffusion to create antibodies targeting specific regions of bacterial and viral proteins, including those used by SARS-CoV-2. The designed antibodies were then experimentally tested for their ability to bind to the intended targets. The successful antibodies did not bind particularly strongly and required modifications for therapeutic use, but the study demonstrates the potential for AI-guided design of antibody drugs.

The study also demonstrates the successful de novo design of single-domain antibodies (VHHs) using a fine-tuned RF diffusion network. These VHHs can specifically bind to user-specified epitopes, offering a novel approach to antibody discovery that bypasses traditional time-consuming methods. Experimental validation confirmed that the designed VHHs successfully bind to disease-relevant epitopes, with the cryo-EM structure closely matching the design model.

The application of AI in antibody design is opening new avenues for the development of antibody therapeutics. With its capacity to reduce discovery timelines and increase success rates, AI is set to play a transformative role in the future of antibody drug discovery.

Related posts:

VIRTUALMAN Breaking Barriers in Drug Development with AI Solutions Sanofi Goes “All In” to Lead the AI Revolution in Pharmaceutical Industry Billion-Dollar AI Deals Powering Antibody Design, Biologics, and Next-Gen Therapies AI-Powered Blood Test Achieves 98% Accuracy in Early Breast Cancer Detection

Author

Oscar Wu