I have blogged previously about us adding the NMRShiftDB database of chemical structures into the ChemSpider database. That blog focused on the performance of NMR prediction algorithms. ACD/Labs has just posted their performance validation of their carbon-13 NMR prediction algorithm on the NMRShiftDB dataset (yes, I’m one of the authors). It makes for very interesting reading. An intro and a link to the review is posted on Ryan’s NMR blog.
The impressive statistic is the average deviation of 1.59 ppm comparing over 214,000 experimental to predicted chemical shifts. Believe me, having been involved with the development of NMR prediction software for over a decade, this is an amazing result. I truly take my hat off to the development team involved with this work. As an NMR spectroscopist myself my vision has been that an average deviation of 1.5ppm would be difficult to achieve over a diverse dataset. This result demonstrates I may be proven wrong shortly! I like being wrong under such circumstances….it’s a cause for celebration.