Comments on: New ChemSpider Functionality at ACS Spring 2010 Part 2 NMR Prediction

By: wolfgang robien

wolfgang robien — Sun, 21 Mar 2010 17:35:53 +0000

To HKO: You are right – I think your statement is with respect to my first comment and I assume you have posted it before you have seen my second comment ! The second comment deals exclusively with spectrum prediction.

Your proposal to get knowledge about existing electronic spectral data:

see http://nmrpredict.orc.univie.ac.at/csearchlite/inchikey.htm
This was announced on November 19th, 2007 (nearly 2,5 years ago !!!!!!) – more to come – stay tuned!

Wolfgang

By: hko

hko — Sun, 21 Mar 2010 12:20:17 +0000

To Wolfgang: Both cited applications are dereplication tools which need known (!) nmr shifts. However in my opinion, the intention is to deliver known or predicted nmr shifts. Therefore it would be at least useful for chemspider users to get information with respect to the selected structure, if there exist known nmr data available in one of your databases.

By: Wolfgang Robien

Wolfgang Robien — Sat, 20 Mar 2010 18:41:23 +0000

Its great to have NMR-Prediction online – there are 2 questions: 1) Whats the price 2) Whats the quality

ad 1) Its free – GREAT !

ad 2) Thats easy to answer, lets do a few tests:

a) Use a simple chlorinated compound e.g. 6-chloro-2-methyl-hex-2-ene
see here: http://www.chemspider.com/Chemical-Structure.14696229.html
Call this entry, display the C-NMR Spectrum ( I got 131.7/127.8/43.9/2×29.6/2×20.3 ppm) – the two -ch2- are very similar, but the cis- and trans- methyls differ usually by about 8ppm, they are given as 2×20.3 ppm

Lets check how other systems perform – lets test CSEARCH:

http://nmrpredict.orc.univie.ac.at/c13robot/robot.php

draw the structure, fill out the data, fill into the first box holding the lines a value of 399.0 – this will cause the program to do a spectrum prediction – thats it, now submit it ….. you will get the results back soon to your mailbox. (requirement is, that your email has been registered IN ADVANCE !)

b) you can repeat this experiment with any terminal di-methyl-’ene’ —> you will learn that NMRShiftDB cant handle the most simple type of cis/trans-isomerism. (except the identical compound is in its database, but then it is no prediction furthermore !)

In order to remove the decision between FREE-OF-CHARGE and scientifically POOR ( state-of-the-art of ca. 1978 ! ) versus NOT FREE-OF-CHARGE, but STATE-OF-THE-ART 2010 the service at http://nmrpredict.orc.univie.ac.at/c13robot/robot.php has been installed – 3 predictions per day are available. I reserve the right to remove this service whenever I want. Email-registration IN ADVANCE necessary !

Stay tuned, more examples will come ! wolfgang.robien(at)univie.ac.at

By: Wolfgang Robien

Wolfgang Robien — Tue, 16 Mar 2010 21:05:32 +0000

At http://nmrpredict.orc.univie.ac.at/case/propose.php you find a structure dereplication engine based on approx. 20,000,000 compounds from PUBCHEM. Spectra have been predicted using the CSEARCH NN-approach leading to approx. 3,000,000,000 structure-spectra pairs ( for details see http://nmrpredict.orc.univie.ac.at/csearch_summary/strpro.html ) – best search time is ca. 0.4 sec / average search time tested on 500,000 examples is below 1 second. Isn’t that bad for 250 GB of data ) Result is linked to the INCHIKEY-pages ( see http://nmrpredict.orc.univie.ac.at/csearchlite/inchikey.htm ) – you immediately know that there are experimental data in either NMRPREDICT, SPECINFO, KNOWITALL or CSEARCH ( including upcoming data not yet released ) available.

If you prefer a system able to rank the hitlist, then you are right at http://nmrpredict.orc.univie.ac.at/identify/ ; this system is based an some 16 millions of structures again from the PUBCHEM-collection. Search time here is about 5-8 seconds.

Access is free, if you like those systems drop me a line ! Wolfgang Robien