v0.4.4

SingleCellProjections v0.4.4

Diff since v0.4.3

Fixed

• Compat with SCTransform v0.3.
• Add Muon compat (used for package extension).

v0.4.3

SingleCellProjections v0.4.3

Diff since v0.4.2

Added

• Package extension for Muon.jl that allows loading data from .h5ad files using the functions create_datamatrix, create_var and create_obs.

Fixed

• Deprecated old loadh5ad function that only supported some versions of the .h5ad format.

Merged pull requests:

• Lazy dep on Muon.jl for loading h5ad files (#24) (@rasmushenningsson)

v0.4.2

SingleCellProjections v0.4.2

Diff since v0.4.1

Fixed

• Fix bug when merging samples with different sets of features, which sometimes could cause unsorted row values and thus sparse matrices with invalid internal representation.
• project now handles kwargs that are only needed for some of the projection steps (by simply ignoring them when not needed).

v0.4.1

SingleCellProjections v0.4.1

Diff since v0.4.0

Added

• local_outlier_factor, local_outlier_factor_table, local_outlier_factor_projection, local_outlier_factor_projection_table

Fixed

• Updated tutorial to reflect changes in SingleCellProjections v0.4.

v0.4.0

SingleCellProjections v0.4.0

Diff since v0.3.9

Breaking

• DataMatrix will now always use the first column of var/obs annotations as ID. (Multiple ID columns are no longer supported.)
• load_counts - The default obs ID column name is now "cell_id" (was "id" before).
• load10x - default to using only first column (id) as unique identifier. Specify e.g. var_id="var_id"=>["id", "feature_type"] to merge multiple columns to create the ID.
• load10x - default to using first column (barcode) as unique identifier.
• load10x - no longer supports copy_obs_col kwarg.
• set_var_id_cols! is replaced with set_var_id_col! (since there is only one ID column).
• set_obs_id_cols! is replaced with set_obs_id_col! (since there is only one ID column).
• Update to SCTransform 0.2, which handles logcellcounts better when there are multiple modalities (e.g. RNA and antibody counts) present in the data.

Added

• var_counts_fraction - Just like var_counts_fraction!, but not modifying the object in place.
• var_counts_sum and var_counts_sum! - For summing over selected variables. Useful for counting e.g. total RNA expression and finding number of expressed features.
• Added support for using external annotations where applicable (filter, transforms, normalization, statistical tests, var_counts_fraction!, var_counts_sum!)
• Added experimental (thus yet unexported) Annotations struct, that wraps a DataFrame with IDs in the first column, and ensures that ID remain when accessing columns. (So that the resulting object can be leftjoined to data.obs/data.var.)

Fixed

• Add compat for weakdeps (UMAP, TSne, PrincipalMomentAnalysis).
• SVDModel now only stores U and S since V is not needed for projection.

Merged pull requests:

• Breaking changes for 0.4 - simplified IDs, support for external annotations and SCTransform update (#20) (@rasmushenningsson)

v0.3.9

SingleCellProjections v0.3.9

Diff since v0.3.8

• Relax === to == when comparing some models. (This fixes a bug occurring when a model is saved to disk using e.g. JLD2 and then loaded again.)

v0.3.8

SingleCellProjections v0.3.8

Diff since v0.3.7

svd, force_layout and pma now supports seed kwarg. To use it, StableRNGs must be loaded.

v0.3.7

SingleCellProjections v0.3.7

Diff since v0.3.6

load_counts - You can now pass a single filename to load a single file (previously arrays were required, but the error message was confusing).

v0.3.6

SingleCellProjections v0.3.6

Diff since v0.3.5

Added

• local_outlier_factor! - Compute the Local Outlier Factor (LOF) for each observation in a DataMatrix. Supports finding neighbors in a low dimensional space (e.g. after PCA or UMAP), but computing distances in a high dimensional space (e.g. after normalization).
• local_outlier_factor_projection! - Compute the Local Outlier Factor (LOF) for each observation in a DataMatrix. Only points in the base data set are considered as neighbors.

Fixed

• knn_adjacency_matrix - kwarg make_symmetric must now be specified by the caller.

Merged pull requests:

• Local Outlier Factor (#15) (@rasmushenningsson)

v0.3.5

SingleCellProjections v0.3.5

Diff since v0.3.3

Merged pull requests:

• Fix typo in tutorial.md (#11) (@M-PERSIC)