Software Open Access
Colin Megill; Bruce Martin; Charlotte Weaver; Sidney Bell; Severiano Badajoz; Matt Weiden; Justin Kiggins; Jeremy Freeman; fionagriffin; bmccandless; Marcus Kinsella; Snyk bot; Prete; Philipp A.; Marcio von Muhlen; James Taylor; Isaac Virshup; Gökçen Eraslan; GenevieveHaliburton; Alex Wolf
Summary: 0.15 resolves incompatibilities with the current anndata/scanpy releases, improves ontology-driven label input in user annotations, moves annotations to a supported feature, and resolves a number of miscellaneous bugs and UI issues.
User annotations feature has moved to a fully supported feature, and is no longer experimental (nb. command line params have changed, and annotations are enabled by default - see documentation for more information).
Experimental support for ontology-driven annotation/label input was further enhanced. Users can now pick from an ontology or define their own label in a single dialog, and ontologies are available for both label and category creation and editing.
Code coverage reports added to the github repo CI pipeline
Significant additions to unit & smoke tests and configuration testing in the CI pipeline
Added experimental support computed re-embedding on user-selected cell sets.
Better handling of annotations with very large number of categorical values.
Resolved incompatibility with annata 0.7 and remove pinned Python dependencies for anndata, h5py and tables modules. This resolves several compatibility issues with H5AD files. #1157, #1202
Reset UI replaced with Subset and Unsubset (#1183), and unsubset no longer deletes genes in the right side bar (fixes #1171).
Work in preparation for a web-hosted cellxgene (not yet available or complete), including:
Incremental loading of data on the front end
Refactoring the back-end to support a cloud-native data format built upon tiledb.
#1140, #1163, #1200
Variety of small UI improvements throughout the application, including more readable data labels on the main graph (#1141, #1160), better handling of long strings (#1206), improved enable/disable state on buttons, etc.
Datasets with more than 32 annotations would occasionally cause the crossfilter to get confused, and selection state to be lost (#1162)
Undo/redo was not working correctly with color-by-annotation and other UI actions (#1165)