Join ACD/Labs part of Revvity Signals at SMASH 2026 to explore how automated structure verification is helping real research organizations accelerate confident decisions and reduce manual review. You’ll also get a look at how we’re helping advance the use of AI for faster structure elucidation.
Come by our poster Automated Structure Verification: Experiences from the Field and visit our booth to discover how we can help modernize your NMR data analysis workflows.
Poster Presentation
Automated Structure Verification: Experiences From the Field
Dimitris Argyropoulos; NMR Business Manager; ACD/Labs
Dimitris Argyropoulos, Sarah Srokosz, Sergey Golotvin, Rostislav Pol, Uliana Bortnik, and Maxim Kisko
Advanced Chemistry Development, Inc. (ACD/Labs), Toronto, ON, Canada
Automated structure verification (ASV) has long been a kind of holy grail for NMR spectroscopists. Having a fast, expert, unbiased, and reliable computer system analyze the data entered and return a good or bad rating would be an ideal way to increase productivity and reliability and decrease errors. Thanks to the vast improvement in NMR processing algorithms, better accuracy of NMR predictions, and instrumental evolution that allowed for better quality spectra, this idea has become more of a reality over the past couple of decades1.
We have worked with several pharmaceutical organizations, with user groups ranging in size from a few individuals to more than 100 users, to deploy and fine tune ASV systems in their environments. In this presentation, we will discuss the lessons learned from these experiences and solutions to common problems encountered.
The problematic outcomes (false positives and false negatives) can be separated in three categories: sample and preparation oversights, NMR experimental deficiencies, and algorithm shortcomings. Sample and preparation oversights include very dilute and/or impure samples, broad signals, high water content (especially when DMSO-d6 is the solvent), and presence of rotamers. NMR deficiencies occur when the experiments set up are not adequate to give a robust answer or when the selected experiments are setup incorrectly. Such cases include recording a single 1D 1H for a sample with a complex proposed structure or not taking the necessary precautions regarding the presence of heteroatoms and special functional groups. Algorithm shortcomings comprise mainly of challenging peak picking and integration in complex spectra and prediction inaccuracies due to a lack of suitable training data.
Interestingly, almost all the NMR and algorithmic problems can be resolved by adjustments of the processing and analysis parameters, making the sample oversights the only issue that must be addressed before recording the spectra. We will present the specifics of these solutions and then take a step back to examine the question of whether such a system can currently provide enough benefit to justify the investment.
1. Golotvin, S.S., Vodopianov, E., Pol, R., Lefebvre, B.A., Williams, A.J., Rutkowske, R.D. and Spitzer, T.D. (2007), Automated structure verification based on a combination of 1D 1H NMR and 2D 1H–13C HSQC spectra. Magn. Reson. Chem., 45: 803-813. https://doi.org/10.1002/mrc.2034
Learn More