Anyone who has ever struggled to fold a roadmap should have an extra measure of respect for protein molecules, which fold up all on their own and practically put themselves away in the glove box. Protein folding is so remarkably efficient that it has been called a paradox. Thirty years ago Cyrus Levinthal pointed out that a typical protein molecule has so many possible configurations that it would need eons to explore all of them and find the best shape; yet proteins fold in seconds.
Looking at the tangled loops and coils of a folded protein, you might imagine that the arrangement is haphazard—like a randomly crumpled map rather than a properly folded one—but in fact every twist and turn is precisely specified. Chemically, a protein is a linear polymer, a sequence of the smaller molecules called amino acids, which are joined end to end like pop-beads. The sequence of amino acids is the only information about the protein encoded in the genes, but the protein can do its job only if the one-dimensional chain of amino acids folds into the correct three-dimensional structure. Apparently the sequence alone is enough to guide the folding. If two protein molecules have the same sequence, they fold up into the same shape.
One way to gain a better appreciation of the protein molecule's knack for folding is to simulate it with a computer program. The most detailed simulations track the motion of every atom and try to reproduce all the chemistry and physics going on in the system. The ultimate goal is to predict the native structure of the protein based on nothing more than the sequence of amino acids. Unfortunately, that goal is a distant one. The models require hours of computer time just to simulate a few picoseconds of molecular dynamics.
I have been exploring a protein model at the other end of the complexity scale—a minimalist model, where every aspect of the simulation is reduced to its simplest possible form. A model so abstract cannot reveal anything about the structure of particular protein molecules—it cannot show how insulin or myoglobin folds—but it may offer clues to some general principles of protein folding. For example, one might hope to learn what kinds of amino acid sequences lead to a stable and compact molecule.
The great advantage of a really simple model is that you can solve it exactly, at least for short chains of amino acids. You can examine every possible folding of every possible sequence, picking out the ones of interest. You can know with certainty which configurations have the most favorable properties.
Another advantage of a minimalist model is that you don't have to be an expert in protein chemistry or molecular dynamics to play with it. A curious amateur can write a rudimentary program in a few days or weeks, and run it on commonly available machinery. Indeed, the simplified protein structures are so well suited to the needs of the amateur that I am tempted to call them amteins—they're not quite ready to turn pro yet. However, I have been persuaded to choose a name slightly less facetious, and so I shall call them prototeins.