gmattedi / chemicalspace Goto Github PK

ChemicalSpace.__init__ takes in a featurizer function and a distance metric.

Allow passing args and kwargs to from_smi and from_sdf to allow setting the needed __init__ parameters from there as well

Title

Optionally allow to hash molecules by their InChI Key and index

Title

Clustering methods that support or not a n_clusters argument are not handled in a very elegant way

To reduce dependency on local environment

Title

Returning a ChemicalSpace layer object with factory and passing hardcoded arguments is likely to cause issues if a derived layer takes in other attributes that were not specificed. Find a way to relax this

Consider parallelizing molecule hashing

Title

Title

Title

Implement Sphere Exclusion Clustering as clustering method

Investigate

Consider adding functionality to align the representation of the molecules of a chemical space to a 2D template

indexing via __getitem__ should return a ChemicalSpace object, and not unpack the result as a tuple of mols, indices and scores

As things are now, features are cached.
But if one were to split a ChemicalSpace by slicing/clustering/etc, then features would need to be recalculated each time

Title

Add a ChemicalSpace.draw option, or equivalent, to render the molecules in the space

Or even 3.6