There have been a number of discussions flying around blogspace recently about NMR Prediction Algorithms. Very interesting to watch and primarily focused on C13 NMR prediction.

Last night I finished up proofing a chapter written with a friend and colleague, Gary Martin, on the applications of N15 NMR in Alkaloid Chemistry for a book to be published by Wiley in the near future (Modern Alkaloids, Structure, Isolation, Synthesis and Biology, Fattorusso, E. / Taglialatela-Scafati, O. (eds.)).

Gary has certainly been my mentor on N15 NMR in recent years as well as providing me with weekly updates to tantalize my eyes..he is a PHENOMENAL photographer.

The book cover is shown below…it will likely become a definitive reference text for alkaloids chemists.

Modern Alkaloids Book


























During the process of writing this book chapter we had the opportunity to test the performance of N15 NMR prediction algorithms using the program I have access to, ACD/NNMR Predictor.

Suffice to say, the performance was excellent. A plot of observed vs. calculated 15N chemical shifts for forty-nine 15N chemical shifts from the review article is shown below. These chemical shifts and associated compounds are not contained within the training set.

N15 Regression



Regression analysis delivered: R2 = 0.987. Further details of the analysis will of course be in the book (I don’t get any royalties honestly!).

The results are in alignment with those we obtained when examining the performance of P-31 NMR prediction for a book I co-authored “Practical Interpretation of P-31 NMR Spectra and Computer-Assisted Structure Verification“.

XNMR prediction is not as refined as C13 NMR or even H1 NMR prediction for sure…the datasets are MUCH smaller than the more common nuclei and the algorithms are not yet optimal. However, XNMR prediction certainly provides value…and if nothing else, keep me up at night with the other passionate NMR jocks in the world clicking on the keyboard…

Stumble it!

3 Responses to “N15 NMR, Performance Validation of Prediction Algorithms based on Alkaloids”

  1. Ryan's Blog on NMR Software says:

    N15 NMR Prediction Validation and Lighthouses

    Tony at the ChemSpider Blog points us to two gems. A peek at an 15N NMR prediction validation study that was conducted during the writing of a book chapter to be published by Gary Martin in the book entitled Modern

  2. Egon Willighagen says:

    Please provide a list of InChI’s for the test compounds. I can then check how good the NMRShiftDB does. The compound on the cover looks like InChI=1/C11H13F2N5O/c12-7-4-8(18-9(7)13)10(19)15-3-1-2-6-5-16-11(14)17-6/h4-5,18H,1-3H2,(H,15,19)(H3,14,16,17).
    The NMRShiftDB does not contain many 15N spectra, and actually fails to assign shifts to two of the five nitrogen nucleis in this compound.

    What about 195Pt?

    Seriously, 13C, 1H and 15N are quite often used, but with current the current NMR apparatus other nuclei are becoming more easily too… what do you think are the odds that we’ll measure the NMR for all nuclei in the molecule we think we have made?

  3. Antony Williams says:

    Egon, I’ll be assembling all of the structures from this review article shortly to use as test submissions to the deposition system of ChemSpider. When that is done then I’ll be able to provide you with the set of InChI codes. It’ll be a few weeks…apologies but many other distractions right now.

    In regards to 195Pt prediction …which I think is your question…I’m not aware of any prediction programs out there. I have measured 195Pt spectra myself a few years ago…great nucleus. It was during a time of messing around with 125Te, 77Se and 51V. I really enjoyed playing with the exotic nuclei. There have been reports on 195Pt shifts and deriving prediction equations for a small related set would likely be very easy. However, for general chemical space for all variants of Pt compounds would be challenging since there is likely a fairly small training set. A good reference text by the way is “R.K. Harris and B.E. Mann in NMR and the Periodic Table, Academic Press, London, UK, 1978. ISBN: 0123276500 9780123276506″

    In regards to your third question I will comment on that in a separate blog. I definitely have opinions on that!

Leave a Reply