CellRank dev (2022-10-07)#


  • Update CytoTRACE calculation to check infinite values. #761

  • Refactor cellrank.tl.kernels.TransportMapKernel for easier definition of extensions. #764

  • Allow plotting negatively correlated genes in cellrank.pl.lineage_drivers(). #779

  • Refactor kernel module. These changes include: - allow kernel element-wise multiplication - decouple CytoTRACE computation from kernel initialization, see cellrank.tl.kernels.CytoTRACEKernel.compute_cytottrace() - change argument mode to model in cellrank.tl.kernels.VelocityKernel - change argument scheme to similarity in cellrank.tl.kernels.VelocityKernel - remove 'sampling' option in cellrank.tl.kernels.VelocityKernel - distinguish between unidirectional and bidirectional kernels - bidirectional kernel inversion is no longer in-place, returns a new object - add more memory efficient way of copying kernels - improve cellrank.kernels.PrecomputedKernel - abstract kernel-related plotting functions (projection and random walks) - temporarily remove kernel tricks examples - fix various typos in docstrings - fix corner-case bug when computing absorption probabilities - fix not setting transport maps in cellrank.tl.kernels.TransportMapKernel - fix automatic threshold removing 1 row in cellrank.tl.kernels.TransportMapKernel #791

  • Allow local block-diagonal connectivities in cellrank.tl.kernels.TransportMapKernel. Rename last_time_point argument to self_transitions and add conn_weight. self_transitions can also be applied to specific diagonal blocks only. #828

  • Add cellrank.datasets.bone_marrow(). #886

  • Add argument subset_to_serum to cellrank.datasets.reprogramming_schiebinger() to allow downloading the subsetted data. This includes the transition matrix computed with the cellrank.external.WOTKernel. #890

  • Updates kernels/estimators __str__/__repr__ and allows for better method chaining. #896

  • Allow passing connectivities for transition matrix projection. Useful when the kernel is not kNN-based. #930


  • Fix cellrank.external.kernels.WOTKernel saving OTModel and being unable to pickle itself. #748

  • Fix when associating term/macro states assignment when reference is not categorical string. #750

  • Fix color creating in cellrank.tl.kernels.ExperimentalTimeKernel. #784

  • Fix all q-values being NaN if 1 p-value was NaN. Also warn and set to NaN instead of raise when correlations are not in [0, 1]] interval. #835

  • Do not save enums in anndata.AnnData. #842

  • Fix cellrank.tl.Lineage subsetting (non-existent overlapping keys) and refactor the implementation. #861

  • Adds references to the docs #887

  • Fix computing time to absorption to aggregated terminal states. #923

  • Fix initial/terminal points size in random walk. #929

  • This adapts some of our docs for the 2.0 release; it clears up the landing page and introduces some new pages, including the “teams”, “how to cite us”, and “about cellrank” pages. #936

  • This just adds a new concept figure to be used in the updated docs. #937

  • Update to the overview figure #938

  • Import Iterable from collections.abc and not collections. Iterable was removed from collections in Python 3.10 #943

  • Require Python >= 3.8 and fix cellrank.pl.log_odds() colormap type. #949

  • Allow NaN values when determining foreground/background color. #951


  • Ignore pygam deprecation warnings. #798

  • Fix PETSc/SLEPc in CI for Linux. For now, macOS PETSc/SLEPc remains disabled. #850

  • Remove default real-time and pseudotime arguments from cellrank.kernels.ExperimentalTimeKernel and cellrank.kernels.PseudotimeKernel. #894

  • Fix not being able to use minChi in cellrank.estimators.GPCCA.fit(). Also compute 20 Schur vectors by default. #913

  • Prefer plotting macrostates/terminal in a discrete way. #914

  • Use ", " instead of " or " when joining states. #954


  • Add Discourse badge for questions and discussions. #880

  • Adds references to the bibtex file. #883

  • Add furo theme based on scvi-tools. #885

  • Fix RTD build - submodules. #901