Outside of my day-job at Chemistry World, I teach a couple of small-group tutorial classes of first-year chemistry undergraduates. With a new academic year just starting, I’m getting to know my new students, and one of the first things we cover is their familiarity with drawing and interpreting chemical structures.
Ensuring we have a firm grounding (and ideally a neat hand) in chemical notation is worth spending some time on, given the enormous amount of implicit information that can be embedded in even the simplest structural drawings. And like learning syntax and grammar rules in any other language, my students are expected to follow the conventions that have evolved over a couple of centuries of chemists communicating their ideas and conceptions of molecular complexity.
The common language of structural representation allows chemists to rapidly and precisely communicate to each other concepts that would otherwise take huge paragraphs of unwieldy and easily confused descriptive text (particularly as systematic names rapidly get cumbersome). It also allows communication between chemists who don’t necessarily share a spoken language.
Remnants of this cross-cultural communication are embedded in some of the labels we apply to structural elements. The Z and E labels applied to substituted alkenes come from German – zusammen (together) and entgegen (opposite); as do the g (gerade, even) and u (ungerade, uneven) symmetry labels applied to molecular orbitals to describe phase relationships either side of an inversion centre. The rich history of element discovery is preserved in the various languages represented across the periodic table.
But now we face a new chemical communication challenge. How do we translate these concept-laden, heuristically bound scribbles into something that a machine can interpret and manipulate. Human chemists are adept at representing and interpreting chemical structures in a variety of different ways, depending on the concepts that are most relevant to a particular discussion. But the relationship between those different representations of the same structure may not be picked up by a machine without appropriate metadata and programming.
Likewise, computers don’t necessarily need to apply all of the intrinsic assumptions and simplifications that are embedded in our notation of chemical structures – they can deal with sets of atomic coordinates, extensive orbital combinations and electron density distributions in ways that human minds struggle to process or visualise. But for humans to be able to interpret the output of machine-generated structural data, it once again needs to be translated into something that fits into an appropriate conceptual envelope. How that happens will depend on what information we are trying to extract from the data.
As Victoria Atkinson concludes in her feature, the future of digitally-accessible chemical notation is likely to be both more standardised and more personalised at the same time. The key to that outcome is allowing as little information as possible to be lost in translation between different representations, even as we choose which information to prioritise.
