Abstract

There is a growing need to perform NMR analyses faster, and under full automation control [1]. Advances in available hardware have contributed favourably to this objective, and the routine collection of high-resolution NMR data on large compound libraries is now a relatively simple matter. But without reliable software tools to automatically analyse the data the full benefit of this exercise cannot be realised. Furthermore, this analysis may be part of a broader requirement encompassing other analytical steps, access to databases, etc. A modular software solution makes it feasible to incorporate the analyses as a node in a workflow [2] such as Pipeline Pilot, or KNIME. This modularity and automation is dependent on a substantial scripting engine that uses a programming language similar to JavaScript.

How these software scores are used by the laboratory manager may depend on what question needs answering. They could be used to triage the data, relying on the clear passes and fails, but allowing the specialist to only focus on the borderline cases. A cut-off threshold could be set, and all cases below that investigated by a specialist.

We describe here a fully automated QC control stage which should use the available experimental data to perform these determinations without human intervention, and to a reliable level:
   - Purity
   - Solvent impurity
   - Concentration
   - Data consistent with the supplied structural formula
   - Assign the NMR spectrum

Our software attempts to realise all these objectives using one or more of these data types: NMR, MS, and LC.

A critical component is the capacity to deconvolve every peak in an NMR spectrum [3], and automatically determine whether it relates to solute, dissolution solvent, solvent impurity, or unidentified organic impurity. Further analysis is used to determine whether or not a peak is associated with a labile proton. The software also attempts to automatically define the multiplets in the spectrum relating to the analyte.

Purity determination follows this NMR peak classification. LC-UV data may also be used, if available.

The consistency between data and structure typically involves an assessment of the closeness of the extracted experimental data of the solute with the predicted NMR spectrum. Using 1H and 1H-13C HSQC experimental data (if available), tests are performed to assess key indices. Each of these is ascribed a "score" and "significance". If LC-MS data are available then determining whether or not the expected molecular ion cluster is observed is assessed. All these tests are then combined to provide an overall "quality" to the scores and significances. These can be reported in a simple way - red, amber, or green - or in more detail. As part of the analysis the software automatically performs an assignment of the 1H NMR spectrum. This process is often called Automatic Structure Verification (ASV).

Concentration determination relies on a core functionality we have designed that uses the most common NMR quantitation methods. These are mainly reference peaks, and the use of absolute integrals. This functionality may be invoked as part of the batch processing, providing new information that allows these test solutions to be used more readily in biological tests.

Results and examples of these analyses will be presented.

Please, cite this online document as:
Cobas C., Seoane F., Vaz E., Sykora S., Bernstein M., Automated QC with Compound Structure Verification and Quantitation,
   Poster at 53rd ENC, Miami, April 15-20, 2012, DOI: 10.3247/SL4Nmr12.003.

References

[1] Keyes P., Messinger M., Hernandez G., Automated Structure Verification by NMR, Part 2: Return on Investment,
     American Laboratory, February 01, 2012, and references cited therein.
[2] Pipeline Pilot: Accelrys. KNIME.
[3] Cobas C., Seoane F., Dominguez S., Sykora S., A new approach to ... Global Spectral Deconvolution (GSD),
     Spectroscopy Europe, 23(1), 25-30, (2011).

Discussions

Your comments are welcome and will appear here