Scikit-fingerprints

What's Changed

* Small bugfix for MolFromSmilesTransformer

New features

* Lots of new molecular filters: Veber, Xu, REOS, Oprea, GSK, and more
* Dice distance and similarity
* MaxMin split
* Mol loaders: InChI, SDF, FASTA (aminoacids sequences)
* Dataset loaders for LRGB

Fixes and internal changes

* Add mention of Morgan to ECFP docs
* Time benchmark improvements
* Fix best score in hyperparameter search for regression
* Fiq tqdm progress bar for sequential computation

**Full Changelog**: https://github.com/scikit-fingerprints/scikit-fingerprints/compare/v1.9.0...v1.10.0

Breaking changes

- ECFP no longer uses `use_fcfp` argument for pharmacophoric atom types, instead relying on new API

New features

- ElectroShape fingerprint
- expanded pharmacophore to support 2-point and 3-point variants
- pharmacophoric atom invariants
- more molecular filters
- Butina split
- Tanimoto distance and similarity

Fixes and internal changes

- lots of docs fixes
- type and import fixes
- CODEOWNERS file to automatically tag reviewers

**Full Changelog**: https://github.com/scikit-fingerprints/scikit-fingerprints/compare/v1.8.0...v1.9.0

Breaking changes

- MACCS now has 166 columns, with artificial all-zeros column at index 0 removed
- documentation will now get published on release new version, not on `master` branch merge. Only already available modules will be be documented to avoid confusion.
- some modules were moved around, e.g. in `preprocessing`; this should generally not break high-level imports, but may be a breaking change for some use cases

New features

- molecular splitting methods: scaffold and randomized scaffold
- numerous molecular filters
- metrics: added Spearman correlation and more error-resistant AUROC
- slightly more efficient substructure fingerprints
- better error on SMILES parsing
- array caching for fingerprint cross-validation

Fixes and internal changes

- added additional examples to README
- RDKit fingerprint bugfixes
- updated CONTRIBUTING.md, removed CHANGELOG. Now changes are only available through Github releases page.
- refactored tests
- added PR template

**Full Changelog**: https://github.com/scikit-fingerprints/scikit-fingerprints/compare/v1.7.0...v1.8.0

New features and improvements

- OGB MoleculeNet splits available from HuggingFace
- optimized fingerprint pipeline hyperparameter tuning classes
- multioutput metrics
- RDKit 2D descriptors fingerprint

Minor and internal changes

- simplified release based fully on Git tags
- ArXiv paper link

**Full Changelog**: https://github.com/scikit-fingerprints/scikit-fingerprints/compare/v1.6.1...v1.7.0

Breaking changes

- moved loose utility functions into `utils` directory

New features and improvements

- a lot of new docs, all fingerprints are now fully documented
- functions for loading MoleculeNet datasets, hosted on HuggingFace Hub
- molecule standardizer, inspired by MolVS
- add memory usage and fingerprint tuning to benchmarking
- small MHFP and SECFP performance improvement
- update allowed RDKit version

**Full Changelog**: https://github.com/scikit-fingerprints/scikit-fingerprints/compare/v1.5.0...v1.6.1