mlqm9nmr

mlqm9nmr package is a Python-based kernel-ridge-regression (KRR) model trained on the QM9NMR dataset for ¹³C-NMR chemical shift predictions of organic molecules.

Step 1: Install mlqm9nmr

• Requirements: numpy, scipy, matplotlib, os, bz2

• Download and install the package

    git clone git@github.com:moldis-group/mlqm9nmr.git
    pip3 install -e mlqm9nmr

Step 2: Load or generate training set descriptors

Approximate size of the training descriptor files:

1.5 GB    aCM_4.dat
491 MB    aBoB_4.dat
927 MB    aCM_RBF_4.dat
7.0 GB    aBoB_RBF_4.dat

OPTION A: To load precomputed descriptors shared in this repository

• If you have Git LFS, go directly to Step 3
  

• If you want to install Git LFS, use:

sudo apt-get install git-lfs

Then go to Step 3.

OPTION B: To calculate training set descriptors from scratch

• Execute calc_di.py Python file to create the training descriptor file:

  from mlqm9nmr import create_descriptor_file
  
  descriptor    = 'acm_rbf_4' 
  geometry_file = '../mlqm9nmr/data/SI_baseline_geo.xyz.bz2'
  
  create_descriptor_file(geometry_file,descriptor)
  

Step 3: Prediction of ¹³C-NMR chemical shift for new molecules

Prepare the geometry file

• Create an XYZ file at the PM7 level (save it as 'test.xyz')

  18
  bigQM7w_012883
  C     1.03070  -0.07680   0.06770  
  C     2.53800  -0.21440  -0.12550  
  C     2.99750  -0.46340  -1.49170  
  N     3.09380   0.90540  -0.90860  
  C     4.47940   1.20090  -0.50870  
  C     5.01760   2.53370  -1.00430  
  C     4.47490   2.41010   0.41050  
  H     0.59860  -1.07330   0.29480  
  H     0.52630   0.33730  -0.83250  
  H     0.83500   0.60170   0.92380  
  H     3.17550  -0.57150   0.71420  
  H     2.25180  -0.44020  -2.31440  
  H     3.99580  -0.93590  -1.63370  
  H     5.09800   0.43550   0.01500  
  H     4.34280   2.85880  -1.82600  
  H     6.09080   2.33310  -1.20820  
  H     3.60210   3.09770   0.43410  
  H     5.35240   2.60380   1.06330
  

Run the model

• Run the ML model in python3 (example in mlqm9nmr/test folder)

  For OPTION A

  from mlqm9nmr import calc_nmr
  from mlqm9nmr import plot_nmr
  
  filename   = 'test.xyz'
  descriptor = 'acm_rbf_4'
  
  cs = calc_nmr(filename,descriptor,di_path='bz2')
  plot_nmr(cs)
  

  For OPTION B

  from mlqm9nmr import calc_nmr
  from mlqm9nmr import plot_nmr
  
  filename   = 'test.xyz'
  descriptor = 'acm_rbf_4'
  path       = 'aCM_RBF_4.dat'
  
  cs = calc_nmr(filename,descriptor,di_path=path)
  plot_nmr(cs)
  

References

• This work presenting ML models trained on QM9NMR with new descriptors
  Enhancing NMR Shielding Predictions of Atoms-in-Molecules Machine Learning Models with Neighborhood-Informed Representations
  Surajit Das, Raghunathan Ramakrishnan
  J. Chem. Phys. 164 (2026) 044106.

• QM9NMR dataset
  Revving up 13C NMR shielding predictions across chemical space: benchmarks for atoms-in-molecules kernel machine learning with new data for 134 kilo molecules
  Amit Gupta, Sabyasachi Chakraborty, Raghunathan Ramakrishnan
  Mach. learn.: sci. technol. 2 (2021) 035010.

• QM9 dataset
  Quantum chemistry structures and properties of 134 kilo molecules
  Raghunathan Ramakrishnan, Pavlo O Dral, Matthias Rupp, O. Anatole Von Lilienfeld
  Sci. Data 1 (2014) 1-7.