ACD/CNMR Predictor

Technical Information

The following section provides a detailed description of ACD/CNMR Predictor and how it generates a prediction. CNMR Predictor uses an internal DAT file of over 2,351,000 experimental chemical shifts. Each chemical shift value is pre-assigned to a particular fragment. This data DAT file has been generated from recorded literature values for over 186,000 structures that are viewable in the ACD/CNMR DB Add-on plus many more structures. Also, CNMR Predictor uses many different subalgorithms for estimating chemical shifts for those fragments that are not in the internal database of experimental values. Each separate algorithm predicts chemical shifts for a specific class of structures. Both the algorithms and the approaches of their derivation are proprietary to ACD/Labs, but the following should help you understand what we do in general.

When you draw a new structure in ACD/CNMR Predictor, the program automatically "splits" it into a set of unique fragments which are then compared to the fragments from the internal database:

• If a fragment from your structure coincides with a fragment from the database, the program uses its experimental chemical shift to compose the final CNMR spectrum. For such chemical shifts the program shows zero confidence intervals in the CNMR Spectrum Window (i.e., horizontal bars of zero length on each spectral peak).
   
• If some fragments from your structure are not found in the internal database, the program looks for the most similar fragments in the database. First, the program composes sets of fragments from the database which are structurally very similar to the fragments generated from your structure. Second, the program estimates chemical shift values of fragments from your structure using secondary algorithms and compares them to the estimated chemical shift values of fragments selected from the database. This allows the program to narrow down the set of similar fragments from the database. Third, the program calculates the average values of the experimental and estimated chemical shifts left after applying the second criterion. Then the program calculates the resulting chemical shift value using both the estimated chemical shift value of your fragment and the average chemical shift values. The obtained chemical shifts are used to compose the final CNMR spectrum of your structure.

When you create a User Database with your own structures, the program automatically "splits" all of them into a set of unique fragments. As you assign new chemical shifts to the carbon atoms, the program treats these data as an update to its internal database. If you draw a new structure and include the opened User Database in the prediction, the program performs the same actions as described above, but it pays primary attention to the chemical shifts which you have entered in your User Database.