One of the most exciting events we hold at the CCDC are our blind tests of crystal structure prediction (CSP) methods. This one turned out to be the best yet. CSP is a really attractive problem as the challenge is beautifully simply to express, but monumentally difficult to solve. Perhaps this is why the entire CSP community comes together every few years, often pooling resources, to have a crack. It’s a wonderful model that could be applied to all sorts of challenges in chemistry and other areas of science.
Crystal structure of 3,5-dinitrobenzoic acid and Tröger’s base, which was successfully predicted by a number of methods in the blind test
This press release
and this News and Views
article in Nature describe some of the highlights of the latest test and a full paper will be out soon, followed by research papers from each of the groups. This will include one from scientists here at the CCDC. We set out to answer the question ‘Can’t you just look at similar structures and make a guess?’ The answer - Well no, you can’t! But we have been working on a probabilistic approach to generate and rank plausible structures, learning what we can from over 800,000 entries in the CSD. Results are encouraging and may well complement the stunning successes we see with energy based approaches.
Will crystallographers be out of a job? Not just yet. For some of these systems it’s still much easier just to do the experiment!