A problem hidden in a folded shape

Games have a quality that makes them ideal for this kind of learning and slightly suspicious as evidence. The score is unambiguous, the rules never change, and a bad move costs nothing.

So the question sitting under all of it: does any of this help with a real problem?

Proteins perform many jobs in living things. They help digest food, carry oxygen, fight infections, and build cell structures. Each protein begins as a chain of smaller units called amino acids, linked in a particular order.

Many protein chains fold into three-dimensional structures. Structure strongly affects what a protein can do because proteins interact physically with other molecules. But structure is not the whole story. Motion, binding partners, chemical changes, and the cell around the protein can matter too. Some proteins also remain partly disordered instead of settling into one rigid shape.

This sets up a specific prediction problem. Given the amino-acid sequence, can a system predict a three-dimensional structure close to the one experiments would observe?

Experimental methods such as X-ray crystallography, nuclear magnetic resonance, and cryogenic electron microscopy can require substantial laboratory work. Known sequences therefore grew much faster than experimentally determined structures.