MODELS are a scientist’s favourite toy. Playing with mathematical building blocks is a good way to get an idea of the machinery of a real natural phenomenon. But models by definition trade accuracy for simplicity: they take into account only those variables which are deemed essential. Otherwise, they would become too unwieldy—at least for scientists (or deputised graduate students) to tackle by hand. Computers, though, are better than humans at mindless mathematical grunt work. They have allowed modellers to incorporate ever more factors into their equations, making the toys resemble the real deal to an ever greater extent.

There are, however, limits to the mathematical prowess even of computers. This year’s Nobel prize in chemistry was awarded to a trio of researchers who came up with a clever way of circumventing some of them. Martin Karplus, Michael Levitt and Arieh Warshel collectively helped to tame the daunting mathematical complexity involved in simulating chemical reactions.

A good way to think about chemistry is as applied physics. Chemical reactions involve the gyrations of electrons, whose behaviour is well-understood thanks to quantum mechanics. But physicists have the luxury of dealing with particles in isolation, which keeps the mathematics relatively simple. Chemists must deal with the complexity of the real world, with multiple electrons swirling around in several different substances. And applying quantum mechanics to the real world—even with powerful computers—turns out to be a pig of a problem.

Fortunately, there is an alternative. The old physics of Isaac Newton is simpler than the modern quantum variety. Complicated molecules such as enzymes or drugs can be modelled fairly easily. But Newtonian models tell you only how a chemical is built. They cannot say anything about how it will react in the presence of something else.

Dr Karplus led a research group at Harvard University in the early 1970s that built some of the early quantum-mechanical computer simulations. Meanwhile, at the Weizmann Institute of Science in Israel, Drs Levitt and Warshel had been working on a Newtonian programcapable of handling even large biological molecules such as enzymes, say.

When Dr Warshel arrived at Harvard in 1970, after finishing his PhD, he and Dr Karplus began to collaborate. Their key insight was that it is not necessary accurately to simulate the entire chemical. Although many biological molecules are huge, their active sites—the bits that actually perform the exciting reactions—are relatively small. So the two researchers came up with a hybrid model that used quantum mechanics to mimic the interesting bits of a given molecule, and faster, rough-and-ready Newtonian methods to simulate the rest.

Their initial program suffered from various limitations that meant it was useful in only a handful of special cases. But over the next few years, the three researchers broadened their approach to accommodate virtually anything while keeping the requisite maths tractable. They achieved this by lumping groups of atoms far away from the reaction site together and treating them as single, large pseudo-atoms. Later they twigged that it was possible to treat large parts of a molecule as a single, homogeneous lump of stuff without compromising accuracy. Their model got faster still.

As with any model, the laureates' is a simplified version of reality. But at least it is no longer a dumbed-down one. Striving for anything else would, of course, be to miss the point of building a model in the first place. To quote Albert Einstein, a model must be as simple as possible—but no simpler. And in creating one, the three winners also brought chemistry fully into the computer age.

**Science and technology**

Babbage

Babbage