Upload structure (.xyz, .extxyz, .cell, .scf.out, .cif, POSCAR)

About ShiftML

ShiftML is an umbrella term for fast and accurate machine-learning models that predict chemical shieldings in organic solids. ShiftML has been developed jointly between the Laboratory of Magnetic Resonance (LRM) and COSMO laboratories at EPFL.

ShiftML hosts ShiftML3 [1], the latest iteration of ShiftML models. ShiftML3 is based on an ensemble of point–edge–transformer deep-learning models and can predict full chemical shielding tensors or isotropic chemical shieldings with high accuracy.

ShiftML3 is trained on a dataset of 1.4 million chemical shieldings from 14 000 organic crystals and can predict chemical shieldings for a wide range of organic solids. The training set covers the following 12 elements:

  • H
  • C
  • N
  • O
  • S
  • F
  • P
  • Cl
  • Na
  • Ca
  • Mg
  • K

Against hold-out GIPAW-DFT data the model achieves isotropic shielding RMSE of 0.43 ppm for 1H and 2.32 ppm for 13C. See the preprint and the publication.

How to cite ShiftML

If you use the ShiftML web app, please cite the ShiftML3 model paper:

[1] Kellner, M., Holmes, J. B., Rodriguez-Madrid, R., Viscosi, F., Zhang, Y., Emsley, L., & Ceriotti, M. (2025). A deep learning model for chemical shieldings in molecular organic solids including anisotropy. The Journal of Physical Chemistry Letters, 16, 8714–8722.

Please consider also citing the reference in which the training data for ShiftML3 was generated:

[2] Cordova, M., Engel, E. A., Stefaniuk, A., Paruzzo, F., Hofstetter, A., Ceriotti, M., & Emsley, L. (2022). A machine learning model of chemical shifts for chemically and structurally diverse molecular solids. The Journal of Physical Chemistry C, 126(39), 16710–16720.

If you are feeling generous, consider also citing the following references that trace the history of ShiftML:

[3] Engel, E. A., Anelli, A., Hofstetter, A., Paruzzo, F., Emsley, L., & Ceriotti, M. (2019). A Bayesian approach to NMR crystal structure determination. Physical Chemistry Chemical Physics, 21(42), 23385–23400.

[4] Paruzzo, F. M., Hofstetter, A., Musil, F., De, S., Ceriotti, M., & Emsley, L. (2018). Chemical shifts in molecular solids by machine learning. Nature Communications, 9(1), 4501.

ShiftML3 – FAQ

ShiftML3 predictions aren’t identical for magnetically equivalent atoms. Why?

ShiftML3 is built on the Point Edge Transformer (PET) model, which is not perfectly rotationally invariant. This can introduce tiny, random differences for atoms that are magnetically equivalent. We have verified that these fluctuations are minor and do not harm overall accuracy.

Tip — get identical shielding predictions: Average the predictions over all magnetically equivalent atoms.

ShiftML3 shows large errors versus my GIPAW-DFT shieldings. What’s going on?

Chemical-shielding calculations are very sensitive to the code and convergence parameters used. Only compare ShiftML3 to GIPAW-DFT data generated with exactly the same settings as the training set.

Reference inputs for Quantum Espresso with the correct parameters are available in this Zenodo data repository.

I used identical GIPAW-DFT parameters but still see big errors. What now?

Check the model’s uncertainty estimates (committee variance; see “Advanced usage” above). If the uncertainty is several × the element’s test-set RMSE, the prediction is probably unreliable for your structure.

My calculated shieldings don’t correlate with experiment at all. Why?

  1. Validate the baseline. Make sure reliable GIPAW/PBE results exist (or recompute them) and confirm they correlate with experiment. Inaccurate DFT—often the exchange–correlation functional—can be blamed.
  2. Check your structures. If candidate geometries don’t reflect experimental conditions or the inter-atomic potential used to generate structures is poor, both DFT and ML predictions will stray from reality.
Funding
Developed at École polytechnique fédérale de Lausanne (EPFL)

Grant IDs

  • SNSF logo Swiss National Science Foundation (SNSF) 200020_212046
  • NCCR MARVEL logo NCCR MARVEL (SNSF) 182892
  • ERC logo ERC Horizon 2020 101001890 (FIAMMA)

Searching for the old ShiftML-app?: https://www.materialscloud.org/work/tools/shiftml