Calculating chemists

A new approach to the calculation of NMR spectra could help organic chemists identify stereoisomers of small, but complex molecules, much more quickly according to theoretical chemists at Cambridge University.

In the spirit of doing more with less, Jonathan Goodman and colleague Steven Smith of the Unilever Centre for Molecular Science Informatics have developed a new approach to molecular structure determination that should help organic chemists confirm the structure of molecules, and assess how much confidence they should have in their conclusions. New substances are discovered all of the time. But, without definitive crystallographic data, do we really know their molecular structures and stereochemistry? The difference between a correct structure and bad assignment could be very costly for pharmaceutical companies and graduate students alike. Goodman's technique shortens the odds on exactly how much of a risk you are taking when setting out a structure.

Calculating NMR shifts from first principles has become a useful tool over the last decade since ab initio tools were revealed by Bifulco in the early 2000s. The approach has since played a critical role in the stereochemical assignment or reassignment of several natural products, including hexacyclinol, maitotoxin, gloriosaols A and B, obtusallenes V-VII, samoquasine A, and many others.

Shift calculations have also been used to determine or confirm stereochemistry in synthetic products including bicyclic peroxides, epoxides of carene and isohasubanan alkaloids.

"There are many cases where new molecules have been discovered, but their precise structure is uncertain, or even miss-assigned," Goodman explains. "Our new procedure, known as DP4, tells you how much confidence you can have in the results." He adds that for pharma companies, part of the expense of developing new drugs means that companies bet huge amounts of money on their data being accurate. "DP4 quantifies the odds, and so it is possible to make more rational decisions on how to place the bets," Goodman says.

Goodman explains that it is now possible to be more certain of structure using nothing more than a simple and cheap one-dimensional NMR spectra, which means the need to perform more sophisticated and more expensive NMR experiments could be reduced for many laboratories.

The DP4 approaches hinges on GIAO NMR shift calculations. The GIAO, gauge invariant atomic orbital, calculation can be applied to reliably assigning stereochemistry with quantifiable confidence even if only a single set of experimental data is available. The team has tried various approaches to assigning a probability to each candidate structure and tested each on up to 64 possible diastereoisomers of 117 different molecules. They used NMR shifts from fast and inexpensive single-point calculations on molecular mechanics geometries that avoided time-consuming ab initio geometry optimization.

The team's approach is essentially to calculate the shifts for candidate structures (employing a Boltzmann weighted average of the shifts calculated for all low-energy conformers) and then to compare these with experimental data to find the best match. Of course, errors in the experimental data and the calculations mean that there will never be a perfect match and so a probability factor is employed.

"We show that a probability analysis based on the errors in each 13C or 1H shift is significantly more successful at making correct assignments with high confidence than are probabilities based on the correlation coefficient and mean absolute error parameters," the team explains. "Our new probability measure, complements the probabilities obtained from our previously developed CP3 parameter, which applies to the case of assigning a pair of diastereoisomers when one has both experimental data sets."

"There is a huge number of such spectra in the literature," says Goodman, "and we have started to look through them to see if we can identify overconfident assignments from the past. We are also working with people studying new natural products to help confirm their structures." Indeed, the team has already demonstrated proof of principle with DP4 in re-assigning the stereochemistry or structure of 21 natural products that were originally incorrectly assigned in the chemical literature or that required extensive synthesis of diastereoisomers to establish their stereochemistry.

"DP4 deals efficiently with both stereochemical and structural assignment, and we recommend its use when experimental data is available for a single compound," Goodman concludes.

The views represented in this article are solely those of the author and do not necessarily represent those of John Wiley and Sons, Ltd.