Sign in Subscribe
Machine learning

Neural Networks Replicate Proton and C Chemical Shifts

Strategic Insights

— 1 min read

Researchers have developed neural networks that accurately reproduce proton and C chemical shifts obtained from oB97X-D/6-31G* density functional model GIAO calculations. These networks support uncharged, closed-shell molecules comprising H, C, N, O, F, S, Cl, and Br. The development involved training the networks to 2.7 million equilibrium geometry and chemical shift calculations for a diverse collection of organic molecules, including synthetic drugs and natural products.

Key Takeaways:

  • The neural networks show RMS errors of 0.05 ppm (proton) and 0.76 ppm (C) for 601 marine natural products.
  • When using equilibrium geometries from a previously described 'estimated oB97X-D/6-31G*' neural network model, the RMS errors increase to 0.09 ppm (proton) and 1.02 ppm (C) shifts.
  • The neural networks reduce the computational time required for accurate proton and C chemical shifts from tens to hundreds of minutes to just a few seconds per molecule.
  • The models have been tested on 246 natural products, with 45% of C shifts reproducing experimental values within 1 ppm, 73% within 2 ppm, and 86% within 3 ppm.
  • The research was conducted by Philip E. Klunzinger, Thomas Hehre, Bernard J. Deppmeier, William Sean Ohlinger, and Warren J. Hehre.
  • The study was published in the Journal of Organic Chemistry.

Statistics:

  • 2.7 million equilibrium geometry and chemical shift calculations were performed to train the neural networks.
  • 601 marine natural products were used to test the neural networks, with RMS errors of 0.05 ppm (proton) and 0.76 ppm (C).
  • 246 natural products were used for a second assessment of experimental C chemical shifts.
  • 45% of C shifts reproduced experimental values within 1 ppm, 73% within 2 ppm, and 86% within 3 ppm.

Sources:

  • Klunzinger, P. E., et al. "Practical Machine Learning Strategies 4: Using Neural Networks to Replicate Proton and 13C NMR Chemical Shifts Obtained from oB97X-D/6-31G* Density Functional Calculations." The Journal of Organic Chemistry (2025).
  • NewsRx. "New Machine Learning Study Findings Have Been Reported by Philip E. Klunzinger and Colleagues (Practical Machine Learning Strategies 4: Using Neural Networks to Replicate Proton and 13C NMR Chemical Shifts Obtained from oB97X-D/6-31G* Density ...)." Journal of Engineering. August 18, 2025; p 1491.

Read more