ecotyper.discovery_scrna¶

scRNA-seq discovery (EcoTyper_discovery_scRNA.R):

cell-type-specific genes (state_discovery_scRNA_filter_genes.R);
cell-cell correlation matrices (state_discovery_scRNA_distances.R) and NMF on them (state_discovery_NMF.R with discovery_cross_cor);
rank selection;
initial states (state_discovery_initial_plots_scRNA.R);
marker genes and NMF re-discovery on expression (state_discovery_extract_features_scRNA.R);
initial states on expression;
QC filter (state_discovery_first_filter_scRNA.R);
ecotypes (ecotypes_scRNA.R, ecotypes_assign_samples_scRNA.R).

Functions

`assign_samples_scrna`(initial, states, eco, ...)	`ecotypes_assign_samples_scRNA.R`.
`cell_type_specific_genes`(expr, annotation, ...)	Step 1 for all cell types (each R process repeats the same subsampling).
`cross_correlation`(expr, annotation, ...[, ...])
`de_subsample`(annotation, cell_ids)	Annotation of the cells used for the DE test: `set.seed(1234)` then up to 500 cells per cell type, cell types in sorted order.
`discover_ecotypes_scrna`(initial, states, ...)
`do_cv`(data, variable, value)	`doCV` (lib/misc.R): within-group variance, decreasing.
`do_de`(data, variable, value[, ranks])	`doDE(data, variable, value)` (lib/misc.R): Wilcoxon rank-sum test of one group against all other samples.
`extract_features_scrna`(cc, initial_assignment)
`fast_cor`(X)	`HiClimR::fastCor(X)` with its defaults (`nSplit = 1`, `optBLAS = FALSE`): `cor()` between columns, zero-variance columns set to NA.

Classes

`CrossCor`(cell_type, log_data, scaled_filt, cor)
`ScFeatures`(cell_type, full_fcs, top_fcs, ...)
`SpecificGenes`(raw, filtered)

ecotyper.discovery_scrna.do_de(data, variable, value, ranks=None)[source]¶

doDE(data, variable, value) (lib/misc.R): Wilcoxon rank-sum test of one group against all other samples. ranks: row_ranks() of data, reused across groups (same result).

Parameters:

data (DataFrame)
value (str)

Return type:

DataFrame

ecotyper.discovery_scrna.do_cv(data, variable, value)[source]¶

doCV (lib/misc.R): within-group variance, decreasing.

Parameters:

data (DataFrame)
value (str)

Return type:

DataFrame

ecotyper.discovery_scrna.de_subsample(annotation, cell_ids)[source]¶

Annotation of the cells used for the DE test: set.seed(1234) then up to 500 cells per cell type, cell types in sorted order.

Parameters:: annotation (DataFrame)
Return type:: DataFrame

class ecotyper.discovery_scrna.SpecificGenes(raw: 'pd.DataFrame', filtered: 'pd.DataFrame')[source]¶

Bases: object

Parameters:

raw (DataFrame)
filtered (DataFrame)

raw: DataFrame¶

filtered: DataFrame¶

ecotyper.discovery_scrna.cell_type_specific_genes(expr, annotation, cell_types, mode='cell type specific', scale_column=None)[source]¶

Step 1 for all cell types (each R process repeats the same subsampling).

Parameters:

expr (DataFrame)
annotation (DataFrame)

Return type:

dict

ecotyper.discovery_scrna.fast_cor(X)[source]¶

HiClimR::fastCor(X) with its defaults (nSplit = 1, optBLAS = FALSE): cor() between columns, zero-variance columns set to NA.

Parameters:: X (ndarray)
Return type:: ndarray

class ecotyper.discovery_scrna.CrossCor(cell_type: 'str', log_data: 'pd.DataFrame', scaled_filt: 'pd.DataFrame', cor: 'pd.DataFrame', scaled_full: 'pd.DataFrame | None' = None)[source]¶

Bases: object

Parameters:

cell_type (str)
log_data (DataFrame)
scaled_filt (DataFrame)
cor (DataFrame)
scaled_full (DataFrame | None)

cell_type: str¶

log_data: DataFrame¶

scaled_filt: DataFrame¶

cor: DataFrame¶

scaled_full: DataFrame | None = None¶

ecotyper.discovery_scrna.cross_correlation(expr, annotation, cell_type, specific, scale_column=None)[source]¶

Parameters:

expr (DataFrame)
annotation (DataFrame)
cell_type (str)
specific (list | None)

Return type:

CrossCor

class ecotyper.discovery_scrna.ScFeatures(cell_type: 'str', full_fcs: 'pd.DataFrame', top_fcs: 'pd.DataFrame', log_full: 'pd.DataFrame', log_top: 'pd.DataFrame', top_scaled: 'pd.DataFrame', full_scaled: 'pd.DataFrame')[source]¶