"CSEARCH" and "nmrshiftdb.org"

Quality check of nmrshiftdb-data using CSEARCH-Algorithms:

Procedure:

March 10th, 2007:

• Download of the file NmrshiftdbWithSignals.sdf.zip
• Conversion of the data into CSEARCH SD-file format
• Creation of a CSEARCH-database

March 11th, 2007 / lunchtime:

• Application of the usual CSEARCH techniques which are applied to all incoming data and furthermore to all existing data in regular intervals.
• Recalculation of each experimental spectrum using the Neural Network with stereoinformation as implemented into CSEARCH / ca. 20ms calculation time per spectrum:

March 11th, 2007 / afternoon; after 2 hours of work:

Average deviation is now: 2.19ppm

Summary of data quality:

What has been done to the original data:

• Missing double bonds inserted / correction of aromatic systems
• Assignment errors ( e.g. Cinnamic acid derivative at the C=C-C=O system )
• Correction of wrong structures ( e.g. -COOH istead of -OOH, -O-Ac instead of -COO-Me, -CH2-OH instead of -CHO, missing oxygen at acetate )
• Valency at C,N,O,Cl has been corrected
• The number of functional groups decreases during datacorrection, because groups having e.g. 6-valent carbon are not furthermore within the data
  

This small improvement - which is far away from being perfect and/or complete - without the use of one, single literature citation improves the assignment quality by ca. 6,200 ppm !

• In some cases the shift value doesnt fit the structure, because the connectivity table is wrong
• The name has been automatically generated - therefore it perfectly fits the wrong connectivity table
• The molecular formula doesnt fit the connectivity table - because of the signal multiplicities
• Now the structure goes to PUBCHEM - here the connectivity table is corrected, because of a valency check - the same compound has now a fitting molecular formula which also fits the name.
• Now it links back to a spectrum which doesnt fit the structure again .......

But:   Enjoy - it's free !