metadynamics

I’ve been away in Lausanne, Switzerland, to follow a course on metadynamics. Metadynamics is a so called free energy method: it is an algorithm to get the free energy landscape of a system from a computer simulation of molecules. Without going into the details, in metadynamics you push your system  away from the structures it has already explored, more and more when time passes. So your molecules, instead of being trapped into doing always the same thing, explore all the space of conformation that they can get, and in return you also have the free energy of their conformations (that is, their real probability).

Since my Ph.D. I have always been fascinated by free energy landscapes (FEL). A FEL is a map of what a system can do, and with which probability. The higher the energy, the lower the probability: we can say that energy is what it is needed to make the improbable probable. It’s as simple and as deep as that. Once you get a FEL, you have the map of what will happen often, what will happen less, what will not happen at all. You can literally paint what the plausible routes between these states and therefore you also have a map not only of what your system can do, but how can go from doing one thing to another. You also can know where your system will do choices, and be forced to choose one path or another.

You know, there are two ways of doing science (two inextricably intertwined ways): one is vertical and the other one is horizontal. Vertical science is what classically is thought as science: it is digging deeper and deeper for explanation. It goes from one ring to the next towards a deeper (or higher) chain. The classical physicist is the essential vertical scientist.  Horizontal science doesn’t look for explanations: it describes and finds new data. It means harvesting information, classifying it, describing it, mapping it. It creates the territory where the vertical scientist will dig in. Taxonomists, people who sequence genomes, astronomers who map galaxies, they are horizontal scientists.

I am, inside, an horizontal scientist. I want histories, I want networks, I want maps. I want immense collections of data to put in order and to annotate. When I was a kid -I mean, a 7 years old kid- what fascinated me more of living things weren’t the habits of big predators or the colours of bugs: it was taxonomy. I was obsessed with classification. I looked hours and hours to the pages of my grandfather’s zoology books which dealt with that, taking note of the inconsistencies between them (why one book classifies all Orthopterans as a single order and another divides them in Ensiferans and Celiferans?) and struggling to understand why certain classifications were in the first place (why are Lagomorphs not rodents?). And phylogenetics and cladistics are simply orgasmic to me: to have in the classification a true map of relationships between living things, a tree that is also a story, is a shot of heroin to my brain (Why am I not a taxonomist? Good question for another story).

But then, creating a molecular model and simulating it, and getting its energy landscape, is as much as fascinating. It’s a small world you have complete control on. You can map it, you can know its histories, you can see what structures come in and classify them. It has a kind of Platonic existence: it is like a universe that unfolds under your computer screen. It’s like being a kid and knowing that secret place that you, and only you, know, and that you could spend endlessly time within. And that’s what fascinated me in Lausanne last week (along with fondue): practical tools to map an infinity of unseen worlds, to know what their histories will look like. To predict not a single future, but to know what the set of all possible futures will be.

And most importantly, there is a life metaphor in that. I find myself doing again and again the same mistakes. If only I had a metadynamics algorithm to push myself beyond my energy barriers. If only I could experience the paradise that once Henri Poincare` promised us: that everything will happen, infinitely often, no matter how hard it seems, and so we will know everything, without having to worry.