Joe Townsend's PhD and data-driven science

Joe was examined yesterday by Martin Dove (Cambridge) and Henry Rzepa – of course nothing is official but he was given the indication “minor corrections”. So I will congratulate Joe on having to make corrections. I won’t put words into Joe’s mouth but we have discussed the likelihood of his thesis being Openly available (Cambridge does not – yet – have a mandatory requirement).
Joe has pioneered much of the work we have done here – OSCAR, OSCAR-DATA, early natural language processing, eScience, data-driven science. I and others owe a great deal to that. It hasn’t been smooth – in several cases the vision was ahead of the technology or the data. For example it wasn’t then possible to extract chemical reactions out of published papers by robotic means. (We’re making progress in this area). But it was possible to extract crystallography and that’s the basis of CrystalEye. Nick built the extraction technology and Joe has pioneered the data driven-science – is it possible to validate thousands of data automatically by comparing experiment and theory. The answer – for crystallography – is definitely yes. And, somewhat as a surprise to us, he found that the major cause of variance was experiment, not theory. But in one or two cases it has revealed effects due to the method of calculation and that is “new science”. Not world-shaking and probably of the “oh well, we all know that already” sort, but still science hidden in the data.
That’s the same philosophy as Nick has been investigating with the NMR – can we compare high-level calculations with experiment and use it to analyse the variance. If, if we find systematic effects, can these point out science in the experiment or in the theory. More on that later…