Yesterday the world got the sad news that the famous Arecibo radio telescope had finally collapsed from general deterioration, storm damage, and lack of maintenance. But another piece of news also came out that, for me, more than made up for it. It appears, one of the grand challenges of molecular biology has been solved: scientists now have a tool to predict protein folding.
A word about proteins, the molecular building blocks of life. Like gears and levers in a machine, they perform specialized functions due to their shapes and other properties. Made of long chains of smaller molecules, these chains fold up automatically to form a specific shape. Though Nature “just does it,” the folding process is incredibly complex. For many years, despite great advances, it’s been too tough to model with computers.
To get an idea why, a typical protein is made of hundreds to thousands of amino acids. Like a string of magnets, some parts attract each other; others repel. The number of different ways the string can bunch up is truly astronomical. It’s the huge number of combinations that’s overwhelmed traditional computer systems.
Sure, people have been studying proteins already, but for decades they’ve done it the hard way, making them in the lab, then analyzing with X-rays—very expensive and labor-intensive. The value of knowing a protein’s shape means even a tedious method is still worth it.
The new approach comes from DeepMind, the same folks who developed the AlphaGo and AlphaZero game-playing programs that now dominate the ancient game of Go. The new software, called AlphaFold 2, uses a sophisticated form of deep learning to predict how proteins fold.
“Ho hum,” you might say. “Another expensive toy for scientists to play with. How’s that good for anything in the real world?” For one thing, it has the potential to speed research, whether to study diseases, new species, or drug development. It can also make the study of artificial proteins easier.
And this may open a door to sophisticated types of nanotechnology, or engineered, complex molecules built precisely, atom by atom. Technology has been heading this way for years, anyway; but scientists have few tools to actually construct things at that scale. Custom, protein-based nanomachines may serve as an early stepping stone to more sophisticated atom-stacking systems later on.
What can you do with this? Here are a few examples. Dirt-cheap, highly-efficient solar cells that cover buildings. Microscopic medical robots that accurately and rapidly track your health from the inside. Readily-available replacements for aging and worn-out organs. And diamond as an inexpensive construction material.