Refactor ligand-based models into a library-like project

[jaimergp]

Description

This PR aims to refactor @t-kimber's scripts for ligand-based models into a library-consumable infrastructure.

Todos

Library:

• [ ] Bring in featurizers:
  • [ ] Normalize Hash Featurizer for protein
  • [ ] Dominique's kinase fingerprint
• [ ] Bring in models
• [ ] Design Ligand and Protein helper objects for easy handling of raw data and their potential featurizations. Ligand object can subclass openforcefield.topology.Molecule, but I am not sure how we should deal with Protein...
• [ ] Provide DatasetConsumers (or *Builder, *Handler, etc) base class and corresponding subclasses for KINOMEscan, ChEMBL, etc.
• [ ] Add unit tests
• [ ] Make sure docstrings are there and are accurate @t-kimber

Scripts/workflows:

• [ ] Add CLI-enabled scripts mimicking existing scripts
• [ ] Make sure results are consistent between original and refactored codebases.

Status

• [ ] Ready to go

Ligand

	Class	Docstrings	Tests
Morgan	:heavy_check_mark:	:heavy_check_mark:	:heavy_check_mark:
SMILES	:heavy_check_mark:	:heavy_check_mark:	:heavy_check_mark:
Graph	:heavy_check_mark:	:heavy_check_mark:	:heavy_check_mark:

Protein

	Class	Docstrings	Tests
Hash	:heavy_check_mark:	:heavy_check_mark:	:heavy_check_mark:
Composition	:heavy_check_mark:	:heavy_check_mark:	:heavy_check_mark:
Sequence	:heavy_check_mark:	:heavy_check_mark:	:heavy_check_mark:

Models

	Class	Docstrings	Tests
XGB	:x:	:x:	:x:
DNN	:heavy_check_mark:	:heavy_check_mark:	:x:
CNN	:heavy_check_mark:	:heavy_check_mark:	:x:
MPNN	:x:	:x:	:x:

[t-kimber]

For ChEMBL22 workflow:

• [x] add def main() and parameters
• [x] change print to loggings
• [x] change imports
• [x] pip install -e .
• [ ] change naming in MorganFingerprintFeaturizer class to avoid having molecule.molecule.

[jaimergp]

Superseded by #8