Descriptors Calculation

Python API · stjames models · API example

How it works

Calculate molecular and atom-centered descriptors for comparing structures, exploring chemical space, or building models.

Settings

  • Optimize structure(s)?: enabled by default. Optimize with GFN2-xTB before calculating descriptors; turn this off to evaluate the supplied geometry.
  • Solvent: unset by default. A selected solvent applies ALPB implicit solvation during optimization and adds COSMO surface descriptors calculated with GFN2-xTB/CPCM-X.

Notes

Molecular descriptors include two- and three-dimensional properties from Mordred, using Jackson Burns's mordred-community package, plus measures of molecular size, shape, and flexibility. Atom-centered charges use GFN1-xTB CM5 charges. Fukui indices use GFN2-xTB Mulliken charge differences, and the global electrophilicity index uses GFN2-xTB. The charge, Fukui, and electrophilicity calculations are gas-phase evaluations of the resulting geometry, even when a solvent is selected.

Geometry-dependent descriptors describe the supplied or optimized conformation. Mordred descriptors that cannot be evaluated are omitted, so check for missing values when comparing molecules.

After the workflow completes, inspect the molecular and atom-centered descriptors. For a collection of molecules, compare descriptor columns or use the principal-component analysis (PCA) view, and export the results for further analysis. PCA centers and scales descriptor values and excludes columns with missing values or no variation across the collection.

Submission video

Further reading