These antibodies would fight against COVID-19, Ebola and other viruses.

Scientists have shown that generative artificial intelligence (AI) can increase the effectiveness of antibodies against viruses like SARS-CoV-2 and ebolavirus. The technology, which utilizes so-called large language models, is a shortcut through a laborious process that requires isolation of antibody molecules from the blood of people who have recovered from COVID-19.

Peter Kim, a biochemist at Stanford University in California, said the AI-generated antibody designs were reaching information that is “completely, or largely, non-obvious to even the experts in antibody engineering.” Generative AI uses neural networks that can create content based on patterns they have been trained on.

Immune system proteins called antibodies are widely used in the biotechnology industry, and the AI technology could help to unlock drugs for targets that have proved resistant to conventional design approaches. In this case, a protein language model, created by Meta AI researchers based in New York City and a part of tech giant Meta (NASDAQ: META), was utilized by the scientists to propose a limited set of antibody mutations.

It was found that an unexpectedly large number of the model’s suggestions boosted the capacity of antibodies to bind to their intended targets for SARS-CoV-2, ebolavirus, and influenza.

The technology is expected to enhance not just the current antibodies, but also to generate novel ones that can selectively bind to a desired target, as per the researchers’ priorities. AI could also help to design antibody drugs that can latch onto multiple targets, such as a tumor protein and an immune cell that can kill that tumor.

A key challenge in designing completely new antibodies is the dependence on floppy loops in the antibody structure, and researchers are investigating ways of modeling these interactions with AI.

Related Story:  Meta AI Predicts The Shape Of 600 Million Proteins

Photo by CDC on Unsplash