Calculate a lot of features for the sample dataset

[JJanowiak]

which features to be decided. will need to look at MP prediction papers and what's freely and easily available. Try grouping the features into categories based on what they intend to capture, so specific features could be "turned off" to show the functionality of the library.

Will probably split this into multiple issues later.

[Alex-AMC]

Suggested categories for features:

• 2D/3D
• Chemical / Spatial
• Intermolecular / Intramolecular
• Molecular Data/ Crystal Data

[JJanowiak]

2 csvs

1. features calculations SMILES per row, feature per column

2. feature category feature per row, category per column

[JJanowiak]

Contains a very poor image of potential categories: http://datascience.unm.edu/biomed505/Course/Cheminformatics/basic/descs_fingers/molec_descs_fingerprints.htm

categories are:

• Constitutional: derived from atomic composition (e.g. mol weight, how many rings )
• Topological: derived from molecular graph (Random walks etc)
• Geometric: 3D descriptors (mol volume)
• Electrostatic: derived from partial charges
• Quantum-Chemical: derived from molecular electron wave function (thermodynamic properties, orbital energies)

[JJanowiak]

https://www.rdkit.org/docs/source/rdkit.Chem.QED.html https://www.rdkit.org/docs/source/rdkit.Chem.rdMolDescriptors.html https://www.rdkit.org/docs/source/rdkit.Chem.Lipinski.html

[Alex-AMC]

We could use CDK and a python wrapper To calculate a set of descriptors

[Alex-AMC]

SCINE - Molassembler has the ability to do molecular graphs and a bunch of other descriptors. Written in C++ but has a python key bindings available?

[Alex-AMC]

Descriptors to be saved as .CSV

[Alex-AMC]

Calculated descriptors using built in class for calculating all descriptors : rdkit.ML.Descriptors

Left computer over night to calculate 208 descriptors for 274,978 SMILEs.