Release notes
1.4.0
Changes
get_pseudobulk
changes:Default values now do not filter features. For feature filtering check the new functions
filter_by_expr
orfilter_by_prop
.If feature filters are used, it may return more genes than before due to a change of
> min_props
to>= min_props
.Now it returns quality control metrics such as
psbulk_n_cells
,psbulk_counts
andpsbulk_props
.Now
groups_col
accepts take multiple keys.Now
mode
accepts a dictionary of callable functions. The resulting profiles will be stored in.layers
.
swap_layer
now has a new argumentX_layer_key
, a.layers
key where to move and store the original.X
.Pseudobulk and bulk vignettes have been updated to use the PyDESeq2 package
run_consensus
now accepts extra arguments with the new parametterargs
that are passed down todecouple
.Omnipath functions now return resources with sorted indexes and throw a warning if the version is too old.
run_wsum
andrun_wmean
now correctly accept empty null distributions.
Additions
Added
filter_by_expr
feature filtering function from edgeR.Added
filter_by_prop
feature filtering function. In previous versions it was incorporated insideget_pseudobulk
.Added
plot_psbulk_samples
to assess the quality of pseudobulk samples.Added
plot_filter_by_expr
to assess which filtering thresholds to use infilter_by_expr
.Added
plot_filter_by_prop
to assess which filtering thresholds to use infilter_by_prop
.Added
plot_volcano_df
to plot volcano plots from long format dataframes.Added
plot_targets
to plot downstream target genes of a source by their change and weight.Added
get_collectri
to retrieve the CollecTRI gene regulatory network.Added
get_ksn_omnipath
to retrieve the Kinase-Substrate network from omnipath.Added
rank_sources_groups
to identify marker sources (TFs, pathways, etc.) per group of samples/cells.
1.3.4
Changes
get_pseudobulk
now has new arguments:mode
to change how to summarize profiles andskip_checks
to bypass checks.OmniPath functions now accept more organism synonyms.
Bug fixes
Fixed empty text labels bug for
adjustText==0.8
.
1.3.3
Bug fixes
read_gmt
is now properly exported.
1.3.2
Bug fixes
plot_metrics_scatter_cols
now deals with missing sources when comparing nets.
Changes
get_pseudobulk
andget_acts
now have adtype
argument due to futureAnnData
changes.plot_metrics_scatter
andplot_metrics_boxplot
now useGroupBy.mean(numeric_only=True)
.
Additions
Added
swap_layer
function to easily moveadata
layers to.X
.Added
read_gmt
function to read GMT files containing gene sets.
1.3.1
Changes
Omnipath wrappers (
get_resource
,get_dorothea
andget_progeny
) now accept any organism name.
1.3.0
Bug fixes
Fixed change in api from
sklearn.tree
.Forced gene names in
extract
to be inunicode
format.Changed integer format from
int32
toint64
to accommodate larger datasets across methods.
Additions
Added conversion utility function
translate_net
to translate nets across organisms.
1.2.0
Bug fixes
Removed
python <3.10
limitation.Forced
np.float32
type to output ofget_contrast
.Made
summarize_acts
compatible with older versions of numpy.extract_psbulk_inputs
now checks if mat and obs have matching indexes.plot_volcano
now correctly can plot networks with different source names.
Changes
extract
now removes empty samples and features.run_consensus
now follows the same format as other methods, old function is now calledcons
.get_pseudobulk
now checks if input are raw integer counts.plot_volcano
now can plot without subsetting features by a network and can save plots to disk.plot_volcano
now usesadjustText
to better plot text labels.plot_volcano
now can set logFCs and p-value limits for outliers.get_top_targets
now can also work without subsetting features by a network and returns significant adjusted p-values.get_contrast
now can also work without needing to group.udt
andmdt
now check ifskranger
andsklearn
are installed, respectively.get_toy_data
now contains more example TFs.get_top_targets
now returnslogFCs
andpvals
as column names instead oflogFC
andpval
.format_contrast_results
now returns also the adjusted p-value.
Additions
Added
dense_run
util function which runs methods ignoring zeros in the data.Added
plot_violins
andplot_barplot
functions.Added
p_adjust_fdr
util function to correct p-values for FDR.Added
get_ora_df
function to infer ora from lists of genes instead of an input mat.Added
shuffle_net
function to randomize networks.Added benchmarking metrics
metric_auroc
,metric_auprc
,metric_mcauroc
andmetric_mcauprc
.Added
get_toy_benchmark_data
function to generate a toy example for benchmarking.Added
show_metrics
function to show available metrics.Added
benchmark
,format_benchmark_inputs
andget_performances
functions to benchmark methods and nets.Added
plot_metrics_scatter
function to plot the results of running the benchmarking pipeline.Added
plot_metrics_scatter_cols
function to plot the results of running the benchmarking pipeline grouped by two levels.Added
plot_metrics_scatter
function to plot the results of running the benchmarking pipeline.Added
plot_metrics_boxplot
function to plot the distributions of Monte-Carlo benchmarking metrics.
1.1.0
Bug fixes
Fixed
get_pseudobulk
errors.Fixed
get_progeny
to correctly return non duplicate entries.Fixed
run_viper
parallelization error.Fixed
run_ora
to correctly deal with random ties.
Changes
get_dorothea
now returns an ordered dataframe.get_contrast
now prints warnings instead of returning an empty dataframe.
Additions
Added
get_top_targets
util function.Added
format_contrast_results
util function.