scikit-fingerprints 1.18.0

New features

• Added CT4 and Dice distances
• MoleculeACE datasets
• Train-test split function examples in various splitters

Optimizations

• Optimization of bulk distances computation using sparse matrices
• Optimization of conformer generator using ETKDGv3 -> KDG -> DG as fallbacks, with random starting coordinates
• Optimization of test speed by reducing the number of test points for kNN applicability domain checker

Fixes and small modifications

• Simplified distance tests
• Better error messages in tests
• Update Sphinx build
• Add explicit doc convention to ruff.toml
• Many smaller fixes

Full Changelog: v1.17.0...v1.18.0

scikit-fingerprints 1.17.0

New features

• stratified MaxMin split
• added suppress_warnings option to MolToX transformers
• Klekota-Roth speed optimization, up to 10x faster
• filter condition indicators for increased interpretability
• feature names for all fingerprints
• new applicability domain checkers: kNN, ResponseVariableRange, StandardDeviation, TOPKAT, PROB-STD

Fixes and internal changes

• moved from Poetry to uv
• docs hosting moved to ReadTheDocs
• updated README.md
• fixed output flattening in BasePreprocessor
• updated allowed dependencies
• fixed TDC missing imports
• many small typo fixes

Full Changelog: v1.16.0...v1.17.0

scikit-fingerprints 1.16.0

Breaking changes

• removed second feature of count MACCS fingerprint, count of atoms with atomic number >103, as this can never happen with current definitions and is improbable in general

New features

• applicability domain algorithms: bounding box, convex hull, leverage, PCA, distance to centroids, Hotelling T^2 test
• MCS similarity
• TDC datasets
• MAP fingerprint chirality option

Fixes and internal changes

• many bugfixes
• a lot more tests
• CI improvements, particulalry for datasets
• library logo

Full Changelog: v1.15.0...v1.16.0

scikit-fingerprints 1.15.0

New features

• For each already existing distance and similarity function, bulk versions were added

Fixes and internal changes

• Added dependency on numba for distance calculation speedup

Full Changelog: v1.14.0...v1.15.0

scikit-fingerprints 1.14.0

Breaking changes

• dropped support for Python 3.9

New features

• Python 3.13 support
• many new similarities and distances: Simpson, Braun-Blanquet, Fraggle, Harris-Lahey, Kulczynski, McConnaughey, CT4, Rogot-Goldberg, Russell, Sokal-Sneath
• much faster distances
• PubChem split
• expanded run_in_parallel function to run on single element functions

Fixes and internal changes

• Add more explicit Pandoc information
• Updated tests badge in README

Full Changelog: v1.13.1...v1.14.0

scikit-fingerprints 1.13.1

What's Changed

Patch release that fixes feature names, requested in #389.

Full Changelog: v1.13.0...v1.13.1

scikit-fingerprints 1.13.0

Breaking changes

• Removed one redundant feature from count MACCS, duplicate of another feature
• This is the last release supporting Python 3.9, future versions will work for Python 3.10 or newer

New features

• More interpretable feature names for many fingerprints with .get_feature_names_out()
• descriptors module with some commonly used descriptors, currently topological, constitutional, kappa, and charge
• BCUT2D fingerprint
• First 4 tutorials, also included in the documentation
• More documentation, typo fixes

Fixes and internal changes

• Updated Poetry to 2.0.1
• Migrated from black, isort etc. to ruff

New Contributors

• @chmurzak made their first contribution in #330
• @franciszekjob made their first contribution in #344
• @iedylag made their first contribution in #347

Full Changelog: v1.12.0...v1.13.0

scikit-fingerprints 1.12.0

Breaking changes

• MAP4 implementation has been fixed, and new results in molecular classification are much higher than before

New features

• Feature names for Mordred and RDKit2DDescriptors available via standard scikit-learn .get_feature_names_out() API
• Exposed full TQDM kwargs
• Updated E3FP version
• Added valid_only option to MolFromAminoseqTransformer
• Added examples for dataset loading functions
• Added more comprehensive User Guide

Full Changelog: v1.11.0...v1.12.0

scikit-fingerprints 1.11.0

Breaking changes

• Changed a few filter names to always have Filter at the end of class name
• extract_multioutput_pos_proba is now extract_pos_proba and works for both single-task and multitask cases

New features

• scikit-fingerprints paper has been published at SoftwareX! Added paper link
• virtual screening metrics: EF, RIE, BEDROC

Fixes and internal changes

• Docs fixes for splitters
• README example fix

Full Changelog: v1.10.1...v1.11.0

scikit-fingerprints 1.10.1

What's Changed

• Small bugfix for MolFromSmilesTransformer