New AI Boosts Drug Discovery
Scientists at University of Washington and Harvard University have developed Artificial Intelligence (AI) software that can help create proteins. It may be useful for new vaccines, cancer treatments, or even tools for pulling carbon pollution out of the air.
“The proteins we find in nature are amazing molecules, but designed proteins can do so much more,” says research co-leader David Baker in a press release issued by University of Washington. “In this work, we show that machine learning can be used to design proteins with a wide variety of functions.”
The new AI-based approach to protein design is described in a research paper published in Science. It is inspired by the recent advances in AI, neural networks, and deep learning (subsets of AI) that have allowed researchers to create the text-generator GPT-3 and the image generator DALL-E. Both are frequently covered by the press and often hyped. These AI tools generate text and images from prompts.
Proteins are sequences of chemical building blocks called amino acids, which perform critical functions in living organisms. The new AI software starts with a random sequence of amino acids. And it keeps changing the sequence until a final sequence that encodes the desired function is found. These final amino acid sequences encode proteins that can then be manufactured and studied in the laboratory.
The scientists trained their neural networks using information from the Protein Data Bank. This is a large open database of three-dimensional structural data of large biological molecules, such as proteins.
The team developed two approaches for designing proteins with new functions. “The first, dubbed ‘hallucination,’ is akin to DALL-E or other generative A.I. tools that produce new output based on simple prompts,” reads the press release. “The second, dubbed ‘inpainting,’ is analogous to the autocomplete feature found in modern search bars and email clients.”
The scientists “start with the key features we want to see in a new protein, then let the software come up with the rest,” explains researcher Joseph L. Watson.
A story is published in SingularityHub. It explains that much of the protein design effort is centered on specific hotspots embedded in proteins. These hotspots, or functional sites, are central to most of our basic biological processes. And they act as docks for other proteins or drugs.
The scientists used the new AI software to design cancer drugs and vaccines against viruses. Among other designs, the AI software came out with a perfect match for an existing antibody against a common virus. This indicates that the approach is sound and could eventually help produce new drugs and vaccines.
In other tests, the scientists designed functional sites for an enzyme, protein-binding proteins, and proteins that grab onto metal ions. This is related to how we absorb iron and other important metals, and could eventually permit fixing impairments of this important function.
“These are very powerful new approaches, but there is still much room for improvement,” concludes Baker. “Designing high activity enzymes, for example, is still very challenging. But every month our methods just keep getting better! Deep learning transformed protein structure prediction in the past two years, we are now in the midst of a similar transformation of protein design.”
Don't miss a beat! In our Pulse Newsletter, Thrivous curates the most important news on health science and human enhancement, so you can stay informed without wasting time on hype and trivia. It's part of the free Thrivous newsletter. Subscribe now to receive email about human enhancement, nootropics, and geroprotectors, as well as company news and deals.
Scientists led by Yale School of Medicine have developed a new data analysis method to make epigenetic clocks substantially more ... Read the article →