Generative AI can improve antibody therapies against COVID-19: Here's how
New research shows that generative artificial intelligence (AI) can help improve antibody therapies against COVID-19, Ebola, and other viruses. The latest study, published in Nature Biotechnology, is part of an increasing body of research that seeks to employ neural networks similar to those behind the ChatGPT, to design antibody therapies. Let's take a deeper look at the promising aspects of this technology.
'There's intense interest in discovering and engineering antibodies'
Generative AI is a type of AI algorithm that can generate text, images, and other content, based on what it is trained on. Researchers believe this AI system will help discover antibody drugs for targets that have resisted conventional antibody design approaches. "There's intense interest in discovering and engineering antibodies, and how one makes antibodies better," said Peter Kim, from Stanford University in California.
Generating antibodies with useful properties involves "brute-force screening"
Antibodies are protective Y-shaped proteins that bind like a lock-and-key to invading foreign substances such as bacteria and viruses. Antibodies are among the key research areas since they can be engineered to bind to almost any protein, in turn altering their activity. However, creating antibodies with useful properties and building on them involves "a lot of brute-force screening", said Brian Hie, a computational biologist.
The team used neural networks called protein language models
For the study, the researchers used neural networks called protein language models. The team used a model developed by Meta AI. These neural networks are similar to 'large language models' that are central to AI tools such as ChatGPT. The difference is that instead of being trained on large volumes of text, protein language models are trained on several millions of protein sequences.
Researchers used AI to suggest mutations for the antibodies
Scientists usually employ protein language models to build completely new proteins and to estimate the structure of the molecules with high accuracy. On the other hand, Hie's team used the model built by Meta AI to suggest mutations for the antibodies.
The model suggestions improved the target-binding capability of antibodies
What's interesting is that a large portion of the suggestions from the protein language improved the ability of antibodies, against influenza, ebolavirus, and the SARS-CoV-2 virus—which causes COVID-19—to bind to their targets. "Alterations to a therapy approved to treat Ebola and a COVID-19 treatment bettered these molecules' ability to recognize and block the proteins these viruses use to infect cells," explained Nature.
The model is finding information that's "non-obvious" even to experts
Many of the suggested changes to antibodies occur outside the regions of the protein that interact with its target, which is usually the focus of engineering efforts, said Kim. "The model is reaching to information which is completely, or largely, non-obvious to even the experts in antibody engineering," he added.
Generative AI has a number of other potential applications
Generative AI has shown promising results. It could help address G-protein-coupled receptors, which play a role in neurological disorders, heart disease, and other conditions. Generative AI could also help in the design of antibody drugs that can latch onto multiple targets, such as a tumor protein and an immune cell that can kill that tumor, said Surge Biswas, co-founder of Nabla Bio.
