|
|
Living things are protein patterns.
In the human body, about 20,000 different proteins create our various bones and tissues, operate our chemical processes, and create illnesses when proteins go awry.
How a protein molecule is shaped is key to its function and recently, researchers have been using artificial intelligence (AI) to predict proteins’ 3D structures from the strings of amino acids that comprise them.
Now AI has moved on. Thanks to scientists at the University of Washington, it’s begun designing proteins that haven’t existed yet.
Early versions of the AI designed a brand new protein that could bind to and neutralize a flu virus and a toxic bacterium.
All the researchers had to do was show the AI the design of the part of the protein that would bind to the target. The AI scanned a library of possible protein structures and filled in the rest of the blanks to lay out the complete protein in the correct shape.
The latest AI version is more free-wheeling.
Scientists tell the AI what they need – for example, a protein that will bind to a specific target molecule. The AI then generates random proteins and tests the results against a virtual model of the target.
The program discards proteins that don’t work. When it finds one that might, it mutates the shape and composition over and over until it dead-ends or until it has a protein that meets the goal.
Because this happens at computer speed, tens of thousands of designs can be tried and judged in a matter of days.
The resulting designs then can be built and tested in a lab.
If the AI’s design doesn’t work in actuality, the researchers can give that feedback to the AI and send it back to work.
So far, the program has delivered candidate proteins for new drugs, vaccines, and industrial enzymes.
TRENDPOST: The ability to imagine novel proteins by the thousands promises new approaches to body-friendly disease treatments, vaccines, and industrial processes.
It could create entire new industries.
Inevitably, this new ability will lead some researchers to contemplate designing novel life forms. (The first synthetic life form was created all the way back in 2010.) The necessary ethical considerations and constraints that will need to surround that work will evolve more slowly than the technology and, probably, much more slowly than those life forms themselves.