During the pandemic, intense research activity has begun with the aim of finding effective treatments against COVID-19. The central method has been to develop antibody molecules directed against the virus. Researchers have now shown that artificial intelligence (AI) can be a shortcut to the labor-intensive processes of developing such antibodies, and that AI can be used to enhance the effectiveness of antibodies against viruses. that it Nature News who says this1.
A study was published a few days ago in Nature Biotechnology2 Use AI to design antibodies — not just against viruses, but also against cancers and inflammatory diseases.
Antibody drugs for diseases such as breast cancer and rheumatoid arthritis account for more than $100 billion in global sales each year. Researchers hope that generative AI — neural networks that can generate text, images, and other content based on learned patterns — will speed development and help create drugs for diseases that have hitherto resisted conventional approaches.
Antibodies are the immune system’s most important weapon against infection. These proteins are used in the biotechnology industry, in part because they can be engineered to attach to almost any protein imaginable to manipulate its activity. But producing and improving antibodies with beneficial properties takes a lot of trial and error and takes time.
To see if generative AI tools can make work more efficient, the researchers used neural networks called protein language models, which form the basis of tools like ChatGPT. But rather than being fed large amounts of text, protein language models are trained on tens of millions of protein chains. In the current study, the model was trained on just a few thousand antibody sequences, out of nearly 100 million protein sequences it learned from. A surprisingly high percentage of model suggestions, though, increased the ability of antibodies against SARS-CoV-2, Ebola and influenza viruses to bind to their targets.
These modified antibodies have been used in a well-established and approved treatment for Ebola and covid-19 and have been shown to improve the ability of these molecules to recognize and block the proteins these viruses use to infect cells. The AI model uses information that is unknown or unclear even to those experts in antibody manipulation—they don’t know for sure what’s going on, but the results are amazing.
Not only do researchers want to alter existing antibodies, they also want to use the technology to develop new antibodies that bind to a target of attack, which could be cancerous tumours, neurological diseases, heart disease and many more. So far, researchers have made no progress in this area, but promising studies are under way3-4.
- Callaway E. How generative AI builds better antibodies. Nature News, 04 May 2023. www.nature.com
- Hie BL, Shanker VR, Xu D et al. Efficient Evolution of Human Antibodies from General Protein Language Models Published Online Prior to Print, 2023 Apr 24. Nat Biotechnol. 2023; 10.1038/s41587-023-01763-2. doi: 10.1038/s41587-023-01763-2 DOI
- Shanehsazzadeh A, Bachas S, McPartlon M, et al. Open de novo antibody design with generative AI. bioRxiv, Posted March 29, 2023. www.biorxiv.org
- Eguchi RR, Choe CA, Parekh U, et al. Deep generative design of epitope-specific binding proteins by optimizing underlying morphology. bioRxiv, Posted December 23, 2022. www.biorxiv.org
“Explorer. Unapologetic entrepreneur. Alcohol fanatic. Certified writer. Wannabe tv evangelist. Twitter fanatic. Student. Web scholar. Travel buff.”
Christian’s Lie: New Bull Market or Tug of War?
– This is a large piece of land. We had our own ski slopes – Groruddalen
The defense secretary said he was getting along – turning to questions about taxation