This article is more than 1 year old

Scientists develop AI algorithms to hunt for cancer-fighting antibodies

System identified a candidate that appears more effective than FDA-approved drugs

AI is helping scientists discover new antibodies that could trigger our own immune systems into destroying cancer cells.

The immune system produces antibodies, specialised proteins that can attack foreign cells like bacteria and viruses. Some can attack tumours.

Finding effective antibodies, however, is tricky. Researchers design new antibodies by mutating known examples, growing them in bacterial or yeast cells. These are then tested to see how well they bind with target proteins in lab experiments. The process is repeated multiple times to narrow down the search for the most promising antibodies worth manufacturing.

The screening stage is consuming and expensive, which is where AI algorithms can help. A team of researchers at the University of California San Diego developed a new system that identified an antibody capable of binding 17 times tighter to programmed death ligand 1 (PD-L1), a protein expressed by cancer cells, than atezolizumab, an existing antibody drug recently approved by the US Food and Drug Administration. The researchers hope to develop the new antibody candidate into a drug, we're told. 

"There are millions of mutants of a given antibody and it is impossible to experimentally test all of their binding to an antigen. That is why it is important to develop machine learning methods to accelerate this process," Wei Wang, senior author of the research published in Nature Communications, and professor of Cellular and Molecular Medicine at UC San Diego School of Medicine, explained to The Register

Antigens from cancerous tumours activate the body's immune system to produce antibodies and destroy them. Wang and his colleagues trained an AI model on millions of antibody sequences to predict its ability to bind to a target protein or antigen. 

The resulting AI pipeline is called "RESP" - a term the authors did not define - but which they did suggest is a powerful way to find useful antibodies.

"Our RESP model can predict the binding affinities of a new sequence even if it is not included in the initial screening library. A unique advantage of the RESP model compared to existing AI models is that it [calculates how] confident [its] prediction is, which can greatly help to select a small number of sequences to [test experimentally]," Wang added.

The model screens antibodies more efficiently than traditional computational methods, and scientists can use its predictions to find the most promising new candidates to synthesise and test in lab experiments. AI speeds up the drug discovery process so companies can progress towards clinical trials more quickly. 

"By combining these AI tools, scientists may be able to perform an increasing share of their antibody discovery efforts on a computer instead of at the bench, potentially leading to a faster and less failure-prone discovery process," Wang said in a statement. "There are so many applications to this pipeline, and these findings are really just the beginning."

The team is now using its RESP model to hunt for new antibodies against other antigens, including SARS-CoV-2 to tackle COVID-19. ®

More about


Send us news

Other stories you might like