The confines of chemical space

‘In describing what they do, scientists have by and large bought the metaphor of discovery, and artists that of creation,’ wrote Roald Hoffmann in 1995.¹ But that doesn’t hold for chemistry, he continued - chemists invent and create molecules, which ”puts chemistry very close to the arts’. 

I agree with Hoffmann, which was why I was perturbed to see the idea of chemical creativity challenged by Martin Jansen and J Christian Schön of the Max Planck Institute for Solid State Research in Stuttgart, Germany.² They argued that chemists are not artists painting with a palette of atoms, but explorers wandering a rugged landscape and chancing upon new, fertile valleys. They are not inventors, but discoverers after all. 

This landscape, said Jansen and Schön, is in ‘chemical space’: the universe of all spatial arrangements of atoms. It is a way of saying that, in chemistry, not all things are possible. In fact, most combinations of atoms are unstable and can never be made. In chemical space, these are represented by peaks, ridges, and highlands: their energy is too high for the substances to be viable.  

Chemists must instead seek out the valleys: low-energy configurations, of which there are rather few for any given assortment of elements. Sitting in a dip in the landscape, freedom vanishes - one can’t arbitrarily alter shape or composition without shifting to another dip, which rather limits the role of design in creating a particular molecule. 

Of course, no one questions that a synthetic target has to be stable under the conditions in which it will operate. Yet this doesn’t prevent chemists from exercising the hallmarks of creative work, combining ingredients with a blend of rational design, imagination, flair and intuition. After all, chemical space is too vast for truly systematic mapping. 

Jansen says that ‘we have never questioned that chemists need to be imaginative and creative, not to mention intelligent. What moved us to write our somewhat provocative article was that so many chemists have been wrongly evoking the impression that they were able to design a molecule or a chemical material.’ 

Chemical spaceman 

Now he has expanded on where the notion of exploring chemical space might lead.³ While chemical space might seem an obvious idea in itself, advances in theoretical, computational and combinatorial chemistry are now making it more tangible. Automated synthesis and rapid screening techniques permit experimental searches across large regions of this space by the systematic, combinatorial assembly of many elements or chemical groups. The results have been mixed, however. The approach has marked up some successes in drug development, and is particularly powerful when combined with in vitro selection methods. Yet I’m not aware of any commercially valuable new material that has been identified combinatorially. 

Jansen, however, suggests that there is now no need for blind searching. He points out that chemical synthesis has always been inductive - roughly speaking, it extrapolates from what has been made before. But quantum chemical methods are now accurate enough to map the local contours of the chemical terrain, and are already used to test the stability of potential synthetic targets. 

A good way to explore rugged energy landscapes 

is to use random walks to sniff out downwards gradients, like the Monte Carlo method used for simulated annealing. Jansen and Schön have used this approach to look for candidate synthetic targets for new crystal forms of ionic compounds such as lithium carbonate.⁴ Quick and dirty computational methods can be used to get a general feel of the contours, before switching to high-powered, computationally-expensive techniques for detailed mapping of promising locations. The scientists can then assess promising crystal structures based not only on their stability, but also the plausibility of their packing and symmetry. 

Paradigm shift 

Of course, synthesis generally targets compounds not just because they exist but because they do something. Jansen’s landscape mapping would generate candidates for in silico screening for desired behaviours. But whereas computational methods are pretty good at estimating some physical properties - stiffness, say - others, from catalytic behaviour to superconductivity, are still hard to predict from first principles. There is also the small matter of actually making the stuff.  

So will the ‘deductive paradigm’ based on energy landscapes come to replace the inductive one based on rules of thumb and a traditional, incremental approach? Jansen admits it’s difficult to exhaustively explore even a small region of chemical space, so any paradigm shift is hardly just around the corner. 

I suspect that ultimately the choice will be in part philosophical: do chemists want their options listed and mapped, or will they always prefer to see synthesis as a creative, artistic act?