Check our AQME installation in 2 mins video for a quick installation guide. In a nutshell, AQME and its dependencies are installed as follows:

  1. Install using conda-forge

conda install -c conda-forge aqme
  1. Install using pip:

pip install aqme
  1. Update to the latest version:

pip install aqme --upgrade

Installation of the extra requirements

Extra requirements if xTB or CREST are used (compatible with MacOS and Linux only):

conda install -y -c conda-forge xtb
conda install -y -c conda-forge crest

Extra requirements if CMIN is used with ANI models:

pip install torch torchvision torchani


Known incompatibilities:

  • RDKit cannot be installed through pip install rdkit in Windows when Anaconda prompts are used

Running the tests

Requires the pytest library.

cd path/to/aqme/source/code
cd tests
pytest -v

Features and modules


Module on charge of conformational sampling starting from multiple input types (SMILES, csv, sdf, xyz, etc). Options:

RDKit-based conformational sampling

Faster sampling, suitable especially for unimolecular systems. Options:

  • RDKit standard sampling

  • Systematic Unbounded Multiple Minimum search (SUMM)

  • FullMonte sampling

CREST-based conformational sampling

Slower sampling, suitable for all types of systems (including noncovalent complexes and constrained systems such as transition states)


Module used to refine conformers generated in CSEARCH through new geometry optimizations. Options:

  • xTB (GFN0-xTB, GFN1-xTB, GFN2-xTB, GFN-FF, etc.)

  • ANI (ANI-1x, ANI-1ccx, ANI-2x, etc.)


Generator of input files for QM calculations. Options:

  • Gaussian

  • ORCA

  • pySCF (loading parameters in jupyter notebook)


cclib-based analyzer of output files from multiple QM programs. This module:

  • Separates normally terminated files with no errors, extra imaginary frequencies, duplicates, isomerization to other systems and spin contamination

  • Automatically generates new com files to "fix" the different issues of the calculations with strategies that are optimal for each type of issue (Gaussian and ORCA)

  • Checks that all the calculations are homogeneous (i.e. using the same level of theory, same grid size, same program and version, solvation model, etc)


Descriptor generator from multiple input types such as SMILES, log files, xyz, etc. Descriptors generated with:

  • RDKit descriptors (i.e. number of polar H, number of aromatic rings, etc)

  • xTB (i.e. atomic charges, molecular dipole, solvation energy, etc)

  • QM programs (i.e. descriptors from cclib)