Quantum Torsions

Supporting Data for "Systematic Comparison of Experimental Crystallographic Geometries and Gas-Phase Computed Conformers for Torsion Preferences"

by Dakota Folmsbee (@dlf57) David Koes (@dkoes) and Geoffrey Hutchison (@ghutchis)

All CREST / GFN2 low-energy conformations (over 2.5GB of compressed .tar.bz2 files) are available through Figshare: https://doi.org/10.6084/m9.figshare.21395061.v1

Uncompressed, this data can occupy over 20GB of disk space. GitHub does not permit large files.

tldr: Use qtdg.ipynb and qtdg.txt to sample torsions

We have included qtdg.txt as acyclic torsions SMARTS patterns and qtdg.ipynb as an example RDKit notebook to do torsion sampling / driving using the derived quantum torsion distributions from this work.

In short, generate a uniform random number, and you'll get back a dihedral angle in degrees from 0..360.

These are generated from gauss-fit.txt using inverse-data.ipynb which calculates the cumulative probability distribution from Gaussian fits, normalizes and inverts to generate a set of linear interpolations for dihedral angles (in degrees from 0..360) from uniform random numbers.

Inventory of Files and Directories

• combined-histograms.ipynb
• combined-ring-histograms.ipynb

Jupyter notebooks to generate fits, including fit-figures, comparisons with ETKDG etc. (e.g., cos-fit.txt, cos-ring-fit.txt, gauss-fit.txt, gauss-ring-fit.txt, etc.)

We include a cosine-squared fit which looked better on some patterns but was ultimately worse.

• figures.ipynb
• ring-figures.ipynb

Notebooks to generate figures and ring-figures including histograms of COD, CREST conformers for COD, and combined sets. Uses images from smarts-figures which are generated by https://smarts.plus/

• cod-torsions Text files contain percent of torsions at that degree.
• pqr-torsions, pubchemqc-torsions, and zinc-torsions Text files contain the raw number of torsions at that degree.
• *-crest.tar.bz2 Raw CREST output files in XYZ, usually with associated SDF/mol file