ember_py package

Submodules

ember_py.cli module

ember_py.cli.create_parser()

Creates and returns the ArgumentParser object for the ember toolkit.

ember_py.cli.run_command(parser, args)

Takes a parser instance and parsed args, then executes the correct command.

ember_py.generate_entropy_metrics module

ember_py.generate_entropy_metrics.generate_entropy_metrics(adata, partition_label)

Generate entropy metrics Psi, Psi_blocks (dataframe) and Zeta.

Entropy metrics generated:

  • Psi : Fraction of infromation explained by partition of choice

  • Psi_block : Specificity of infromation to a block

  • Zeta : Speicifcty to a partition/ distance of Psi_blocks distribution from uniform

Parameters:

  • adata (AnnData) – Annotated data object with .obs containing metadata.

  • partition_label (str) – Column in .obs to partition by when calculating entropy metrics.

Returns:

  • Psi (np.Array) – A list of Psi scores (between 0 and 1) corresponding to the selected partition for all genes in .var.

  • Psi_block_df (pd.Dataframe) – A dataframe of Psi_block scores (between 0 and 1) corresponding to the selected partition for all genes in .var. Scores are caluclated for all blocks, each column of the dataframe corresponds to one block.

  • Zeta (np.Array) – A list of Zeta scores (between 0 and 1) corresponding to the selected partition for all genes in .var.

ember_py.generate_entropy_metrics.safe_divide_sparse(numerator, denominator)

Helper frunction performs element-wise division of a matrix by a vector (or array) element-wise while avoiding division-by-zero issues. Supports both SciPy sparse matrices and dense NumPy arrays.

Sparse case:

  • Zeros in the denominator are replaced with inf so the result is 0.

  • Zero entries resulting from division are removed from the sparse representation.

Dense case:

  • Division is performed with suppressed warnings.

  • Positions with zero denominator are set to 0.

ember_py.generate_entropy_metrics.safe_log2_sparse(mat)

Helper function applies an element-wise log2 transform only to the non-zero stored entries in a SciPy CSR sparse matrix. Zeros in a sparse matrix are implicit (not stored in .data) and remain zero in the output.

Any non-finite results from the log transform (e.g., -inf from explicit stored zeros, or NaN from negative values) are replaced with 0. Parameters

ember_py.generate_pvals module

ember_py.generate_pvals.generate_pvals(h5ad_dir, partition_label, entropy_metrics_dir, save_dir, sample_id_col, category_col, condition_col, block_label=None, seed=42, n_iterations=1000, n_cpus=1, Psi_real=None, Psi_block_df_real=None, Zeta_real=None)

Calculate empirical p-values for entropy metrics from permutation test results. This function can be called manually or accessed through light_ember with partition_pvals = True or block_pvals = True.

Manual access useful if wanting to generate p-values for multiple blocks and partitions of interest after initial investigation using entropy metrics.

Integrated access with light_ember is easier if investigating only a partition or a block in a partition.

Entropy metrics generated:

  • Psi : Fraction of infromation explained by partition of choice

  • Psi_block : Specificity of infromation to a block

  • Zeta : Speicifcty to a partition/ distance of Psi_blocks distribution from uniform

Parameters:

  • h5ad_dir (str, Required) – Path to the .h5ad file to process. Data should be log1p and depth normalized before running ember. Remove genes with less than 100 reads before running ember.

  • partition_label (str, Required) – Column in .obs used to partition cells for entropy calculations (e.g., “celltype”, “Genotype”, “Age”). Required to run process. If performing calculation on interaction term, first create a column in .obs concatnating the two columns of interested with a semicolon (:).

  • entropy_metrics_dir (str, Required) – Path to csv with entropy metrics to use for generating pvals.

  • save_dir (str, Required) – Path to directory where results will be saved.

  • sample_id_col (str, Required) – The column in .obs with unique identifiers for each sample or replicate (e.g., ‘sample_id’, ‘mouse_id’).

  • category_col (str, Required) – The column in .obs defining the primary group to balance across in order to generate a balanced sample of the experiment. (e.g., ‘disease_status’, ‘mouse_strain’). Refer to readme for further explanation on how to select category and condition columns. category_col and condition_col are interchangable. If balancing across more than 2 variables, generate interaction terms, create a column in .obs concatnating the two (or more) columns of interested with a semicolon (:). This way balancing can be done across as many variables as desired.

  • condition_col (str, Required) – The column in .obs containing the conditions to balance within each category to generate a balanced sample of the experiment. (e.g., ‘sex’, ‘treatment’). Refer to readme for further explanation on how to select category and condition columns. category_col and condition_col are interchangable. If balancing across more than 2 variables, generate interaction terms, create a column in .obs concatnating the two (or more) columns of interested with a semicolon (:). This way balancing can be done across as many variables as desired.

  • block_label (str, default=None) – Block in partition to calucate p-values for. Default set to None, program will continue generating p-values for only Psi and Zeta.

  • seed (int, default=42) – The random seed for reproducible draws, by default 42.

  • n_iterations (int, default = 1000) – Number of iterations to calulate p-vales. Default set to 1000. Note that doing fewer than 1000 iterations is a good choice to get first pass p-values but for reliable p-values 1000 iterations is recommended. Larger than 1000 will give you more relibale p-values but will increase runtime significantly.

  • n_cpus (int, default=1) – Number of cpus to use to perfrom p-value calculation. Default set to 1 assuming no parallel compute power on local machine. User can input -1 to use all available cpus but one.

  • Psi_real (pd.Series, default=None) – Observed Psi values for each gene. Used by light_ember, not necessary for user use.

  • Psi_block_df_real (pd.Dataframe, default = None) – Observed Psi_block values for all blocks in chosen partition. Used by light_ember, not necessary for user use.

  • Zeta_real (pd.Series, default=None) – Observed Zeta values for each gene. Used by light_ember, not necessary for user use.

Return type:

None

Notes

What to expect inside ‘pvals_entropy_metrics.csv’:

  • gene_name: All genes in .var

  • Psi: Psi scores averaged over n draws (between 0 and 1) generated by light_ember for each gene in .var.

  • Psi p-value: Permutation based empirical p-values for observed Psi scores for each gene in .var.

  • Zeta: Zeta scores averaged over n draws (between 0 and 1) generated by light_ember for each gene in .var.

  • Zeta p-value: Permutation based empirical p-values for observed Zeta scores for each gene in .var.

  • Psi q-value: Multiple testing corrected q-values for Psi scores.

  • Zeta q-value: Multiple testing corrected q-values for Zeta scores.Correction perfromed to include all p-values generated in a single file (Psi and Zeta).

if block_pvals = True and a single block_label is given:

  • psi_block: psi_block scores (between 0 and 1) generated by light_ember for each gene in .var.

  • psi_block p-value: Permutation based empirical p-values for observed psi_block scores for each gene in .var.

  • psi_block q-value: Multiple testing corrected q-values for psi_block scores. Correction perfromed to include all p-values generated in a single file (Psi, psi_block and Zeta).

ember_py.light_ember module

ember_py.light_ember.light_ember(h5ad_dir, partition_label, save_dir, sampling=True, sample_id_col=None, category_col=None, condition_col=None, num_draws=100, save_draws=False, seed=42, partition_pvals=True, block_pvals=False, block_label=None, n_pval_iterations=1000, n_cpus=1)

Runs the ember entropy metrics and p-value generation workflow on an AnnData object.

This function loads an AnnData .h5ad file, optionally performs balanced sampling across replicates, computes entropy metrics for the specified partition, and generates p-values for Psi and Zeta and optionally Psi_block for a block of choice.

Entropy metrics generated:

  • Psi : Fraction of infromation explained by partition of choice

  • Psi_block : Specificity of infromation to a block

  • Zeta : Speicifcty to a partition/ distance of Psi_blocks distribution from uniform

Parameters:

  • h5ad_dir (str, Required) – Path to the .h5ad file to process. Data should be log1p and depth normalized before running ember. Remove genes with less than 100 reads before running ember.

  • partition_label (str, Required) – Column in .obs used to partition cells for entropy calculations (e.g., “celltype”, “Genotype”, “Age”). Required to run process. If performing calculation on interaction term, first create a column in .obs concatnating the two columns of interested with a semicolon (:).

  • save_dir (str, Required) – Path to directory where results will be saved. Required to run process.

  • sampling (bool, default=True) – Whether to perform balanced sampling across replicates before entropy calculation. If True, sample_id_col, category_col, and condition_col must be provided. Sampling should only be False if fast intermediate results are desired or if there are no replicates to sample over. If sampling is set to False but either partition_pvals or block_pvals are set to True then the sampling=False will be overridden as pval generation requires sampling.

  • sample_id_col (str, default = None) – The column in .obs with unique identifiers for each sample or replicate (e.g., ‘sample_id’, ‘mouse_id’).

  • category_col (str, default = None) – The column in .obs defining the primary group to balance across in order to generate a balanced sample of the experiment. (e.g., ‘disease_status’, ‘mouse_strain’). Refer to readme for further explanation on how to select category and condition columns. category_col and condition_col are interchangable. If balancing across more than 2 variables, generate interaction terms, create a column in .obs concatnating the two (or more) columns of interested with a semicolon (:). This way balancing can be done across as many variables as desired.

  • condition_col (str, default = None) – The column in .obs containing the conditions to balance within each categoryto generate a balanced sample of the experiment. (e.g., ‘sex’, ‘treatment’). Refer to readme for further explanation on how to select category and condition columns. category_col and condition_col are interchangable. If balancing across more than 2 variables, generate interaction terms, create a column in .obs concatnating the two (or more) columns of interested with a semicolon (:). This way balancing can be done across as many variables as desired.

  • num_draws (int, default = 100) – The number of balanced subsets to generate, by default 100.

  • save_draws (bool, default=False) – Whether to save intermediate draws to save_dir.

  • seed (int, default = 42) – The random seed for reproducible draws, by default 42.

  • partition_pvals (bool, default=True) – Whether to compute permutation-based p-values for the partition_label. P-values are generated by sampling. If sampling = False and partition_pvals = True, the sampling=False will be overwritten. Calls generate_pavls, which can be called manually after metric generation as well.

  • block_pvals (bool, default=False) – Whether to compute permutation-based p-values for the block_label. P-values are generated by sampling. If sampling = False and block_pvals = True, the sampling=False will be overwritten. Calls generate_pavls, which can be called manually after metric generation as well.

  • block_label (str, default = None) – One value in the .obs column for partition_label to use for block-based permutation tests. Required if block_pvals=True.

  • n_pval_iterations (int, default=1000) – Number of permutations to use for p-value calculation.

  • n_cpus (int, default=1) – Number of CPU cores to use for parallel permutation testing. For this script, performance is I/O-bound and may not improve beyond 4-8 cores.’

Return type:

None

Notes

  • Results are saved to save_dir as CSV files.

  • one csv file with all entropy metrics

  • one csv file in a new Psi_block_df folder with psi block values for all blocks in a partition

  • Separate file for pvals

  • Separate files for each partition

  • Alternate file names depending on sampling on or off.

What to expect inside ‘entropy_metrics.csv’:

  • gene_name: All genes in .var

  • Psi_mean: Psi scores averaged over n draws (between 0 and 1) corresponding to the selected partition for each gene in .var.

  • Psi_std: Standard deviation of Psi scores across n draws corresponding to the selected partition for each gene in .var.

  • Psi_valid_counts: Number of valid Psi scores observed across n draws. Only use genes for downstream analysis that have valid counts=num_draws. If valid counts is not close to num_draws, increase threshold for filtering genes with low reads beforehand(recommended <100 reads, increase as needed).

  • Zeta_mean: Zeta scores averaged over n draws (between 0 and 1) corresponding to the selected partition for each gene in .var.

  • Zeta_std: Standard deviation of Zeta scores across n draws corresponding to the selected partition for each gene in .var.

  • Zeta_valid_counts: Number of valid Psi scores observed across n draws. Only use genes for downstream analysis that have valid counts=num_draws. If valid counts is not close to num_draws, increase threshold for filtering genes with low reads beforehand (recommended <100 reads, increase as needed).

What to expect inside ‘Psi_block_df/’:

  • mean_Psi_block_df.csv : A dataframe of mean Psi_block scores (between 0 and 1) corresponding to the selected partition for each gene in .var. Scores are caluclated for all blocks, each column of the dataframe corresponds to one block.

  • std_Psi_block_df.csv : A dataframe of standard deviations for Psi_block scores corresponding to the selected partition for each gene in .var.Scores are caluclated for all blocks, each column of the dataframe corresponds to one block.

What to expect inside ‘pvals_entropy_metrics.csv’:

  • gene_name: All genes in .var

  • Psi: Psi scores averaged over n draws (between 0 and 1) generated by light_ember for each gene in .var.

  • Psi p-value: Permutation based empirical p-values for observed Psi scores for each gene in .var.

  • Zeta: Zeta scores averaged over n draws (between 0 and 1) generated by light_ember for each gene in .var.

  • Zeta p-value: Permutation based empirical p-values for observed Zeta scores for each gene in .var.

  • Psi q-value: Multiple testing corrected q-values for Psi scores.

  • Zeta q-value: Multiple testing corrected q-values for Zeta scores.Correction perfromed to include all p-values generated in a single file (Psi and Zeta).

If block_pvals = True and a single block_label is given:

  • psi_block: psi_block scores (between 0 and 1) generated by light_ember for each gene in .var.

  • psi_block p-value: Permutation based empirical p-values for observed psi_block scores for each gene in .var.

  • psi_block q-value: Multiple testing corrected q-values for psi_block scores. Correction perfromed to include all p-values generated in a single file (Psi, psi_block and Zeta).

ember_py.plots module

ember_py.plots.plot_block_specificity(partition_label, block_label, pvals_dir, save_dir, highlight_genes=None, q_thresh=0.05, fontsize=18, custom_palette=None)

Generate a psi_block vs. Psi scatter plot to visualize block-specific genes.

This function reads p-value data, colors genes based on their statistical significance for Psi and psi_block scores, and highlights the top genes significant in both metrics. Only interpret genes that are significant by both Psi and psi_block since those are genes that have reliable scores after permutation testing. Allows for custom highlighting of a user-provided gene list. Fontsize and color pallette can be customized.

Parameters:

  • partition_label (str, Required.) – The label for the partition, used in the plot title.

  • block_label (str, Required.) – The label for the block variable (e.g., a cell type or condition).

  • pvals_dir (str, Required.) – Path to the input CSV file containing p-values and scores. The CSV must have gene names as its index column.

  • save_dir (str, Required.) – Path where the output plot image will be saved.

  • highlight_genes (list[str], default=None.) – A list of gene names to highlight and annotate on the plot, by default None.

  • q_thresh (float, float, default = 0.05) – Threshold for q-values. Genes are retained if both ‘Psi q-value’ <= q_thresh and ‘psi_block q-value’ <= q_thresh.

  • fontsize (int, default = 18.) – The base font size for plot labels and text, by default 18.

  • custom_palette (list[str], default=None.) – A list of 6 hex color codes to customize the plot’s color scheme. If None, a default palette is used. Provide list of colors int his order: [‘significant by psi’, ‘significant by psi_block’, ‘highlight genes’, ‘significant by both’, ‘cirlce markers’, ‘circle housekeeping genes’, ‘significant by neither’]

Return type:

None

ember_py.plots.plot_partition_specificity(partition_label, pvals_dir, save_dir, highlight_genes=None, q_thresh=0.05, fontsize=18, custom_palette=None)

Generate a Zeta vs. Psi scatter plot to visualize partition-specific genes.

This function reads p-value data, colors genes based on their statistical significance for Psi and Zeta scores, and highlights top “marker” and “housekeeping” genes. Only interpret genes that are significant by both Psi and Zeta since those are genes that have reliable scores after permutation testing. Allows for custom highlighting of a user-provided gene list. Fontsize and color pallette can be customized.

Parameters:

  • partition_label (str, Required.) – The label for the partition being plotted, used in the plot title.

  • pvals_dir (str, Required.) – Path to the input CSV file containing p-values and scores (Psi, Zeta, q-values). The CSV must have gene names as its index column.

  • save_dir (str, Required.) – Path where the output plot image will be saved.

  • highlight_genes (list[str], default=None.) – A list of gene names to highlight and annotate on the plot, by default None.

  • q_thresh (float, float, default = 0.05) – Threshold for q-values. Genes are retained if both ‘Psi q-value’ <= q_thresh and ‘Zeta q-value’ <= q_thresh.

  • fontsize (int, default=18.) – The base font size for plot labels and text, by default 18.

  • custom_palette (list[str], default=None.) – A list of 7 hex color codes to customize the plot’s color scheme. If None, a default palette is used. Please provide list in this order [‘significant by psi’, ‘significant by zeta’, ‘highlight genes’, ‘significant by both’, ‘cirlce markers’, ‘circle housekeeping genes’, ‘significant by neither’]

Return type:

None

ember_py.plots.plot_psi_blocks(gene_name, partition_label, psi_block_df_dir, save_dir, fontsize=18)

Generates and saves a bar plot of mean psi block values with error bars.

This function reads two CSV files from a specified directory: one for mean psi block values and one for standard deviations. It plots the mean values for a specific gene as a bar plot with corresponding standard deviation error bars. Fontsize can be customized.

Parameters:

  • gene_name (str, Required) – The name of the gene (row) to select and plot from the CSV files.

  • partition_label (str, Required) – The partition label used to find the correct files (e.g., ‘Genotype’).

  • psi_block_df_dir (str, Required) – Path to the directory containing the mean and std CSV files. Files must be named ‘mean_Psi_block_df_{partition_label}.csv’ and ‘std_Psi_block_df_{partition_label}.csv’.

  • save_dir (str, Required) – Path to directory to save the output plot image.

  • fontsize (int, default=18.) – The base font size for plot labels and text, by default 18.

Return type:

None

ember_py.plots.plot_sample_counts(h5ad_dir, save_dir, sample_id_col, category_col, condition_col, fontsize=18)

Generate a bar plot showing the number of unique individuals per category and condition.

This function reads an AnnData object from an .h5ad file in backed mode, calculates the number of unique individuals for each combination of a given category and condition, and visualizes these counts as a grouped bar plot. Fontsize can be customized.

Parameters:

  • h5ad_dir (str, Required) – Path to the input AnnData (.h5ad) file.

  • save_dir (str, Required) – Path to directory to save the output plot image.

  • sample_id_col (str, Required) – The column name in adata.obs that contains unique sample IDs.

  • category_col (str, Required) – The column name to use for the primary categories on the x-axis.

  • condition_col (str, Required) – The column name to use for grouping the bars (hue).

  • fontsize (int, default = 18.) – The base font size for plot labels and text, by default 18.

Return type:

None

ember_py.sample_replicates module

ember_py.sample_replicates.aitchison_mean_and_std(Psi_block_dfs_list)

Compute the Aitchison mean and geometric standard deviation of compositional DataFrames across blocks.

Parameters:

Psi_block_dfs_list (list of pd.DataFrame) – List of compositional matrices (genes x choice_of_partition).

Returns:

  • mean_df (pd.DataFrame) – Aitchison mean of Psi block values.

  • var_df (pd.DataFrame) – geometric std of the Psi block values.

ember_py.sample_replicates.generate_balanced_draws(adata, sample_id_col, category_col, condition_col, num_draws=100, seed=42)

Generate balanced subsets of replicates based on a categorical variable (eg: Genotype, Age, Estrus Stage) and balances draws within each category by a condition variable (eg: Sex, Knockout/Wildtype/Overexpressed, Treated/Control).

For each unique value in category_col, this function creates draws where each unique value from condition_col is represented exactly once. It iterates through the available samples for each category-condition pair to generate diverse subsets across multiple draws.

Parameters:

  • adata (AnnData) – The annotated data object with sample metadata in .obs.

  • sample_id_col (str) – The column in .obs with unique identifiers for each sample or replicate (e.g., ‘sample_id’, ‘mouse_id’).

  • category_col (str) – The column in .obs defining the primary groups to balance within (e.g., ‘disease_status’, ‘mouse_strain’).

  • condition_col (str) – The column in .obs containing the conditions to balance across within each category (e.g., ‘sex’, ‘treatment’).

  • num_draws (int, optional) – The number of balanced subsets to generate, by default 100.

  • seed (int, optional) – The random seed for reproducible draws, by default 42.

Returns:

  • list[list[str]] – A list of draws, where each draw is a list of sample IDs.

  • dict[str, int] – A dictionary tracking the selection count for each sample.

ember_py.top_genes module

ember_py.top_genes.highly_specific_to_block(partition_label, block_label, pvals_dir, save_dir, psi_thresh=0.5, psi_block_thresh=0.5, q_thresh=0.05)

Identifies significant and specific genes from a ember generated p-values/q-values CSV file based on thresholds for Psi, psi_block, and q-values. (Potential marker genes)

This function reads a CSV file containing Psi and psi_block metrics (and their corresponding q-values), filters genes that meet given significance and specificity thresholds, and saves the resulting subset to a new CSV file.

Parameters:

  • pvals_dir (str, Required) – Path to the input CSV file (e.g., ‘pvals_entropy_metrics_Age_E16.5.csv’). The CSV must include the following columns: ‘Psi q-value’, ‘psi_block q-value’, ‘Psi’, and ‘psi_block’.

  • save_dir (str, Required) – Directory where the filtered results CSV will be saved.

  • partition_label (Required) – Name of partition used to generate entropy metrics, used to label saved csv.

  • block_label (Required) – Name of block in partition used to generate entropy metrics, used to label saved csv.

  • psi_thresh (float, default = 0.5) – Threshold for Psi values. Only genes with Psi > psi_thresh are kept.

  • psi_block_thresh (float, Required, default = 0.5) – Threshold for psi_block values. Only genes with psi_block > psi_block_thresh are kept.

  • q_thresh (float, float, default = 0.05) – Threshold for q-values. Genes are retained if both ‘Psi q-value’ <= q_thresh and ‘psi_block q-value’ <= q_thresh.

Returns:

DataFrame containing the significant and specific genes that meet all threshold criteria. Also saved as “highly_specific_genes_by_{partition_label}_{block_label}.csv” in the specified save directory.

Return type:

pd.DataFrame

ember_py.top_genes.highly_specific_to_partition(partition_label, pvals_dir, save_dir, psi_thresh=0.5, zeta_thresh=0.5, q_thresh=0.05)

Identifies significant and specific genes from a ember generated p-values/q-values CSV file based on thresholds for Psi, Zeta, and q-values.

This function reads a CSV file containing Psi and Zeta metrics (and their corresponding q-values), filters genes that meet given significance and specificity thresholds, and saves the resulting subset to a new CSV file.

Parameters:

  • pvals_dir (str, Required) – Path to the input CSV file (e.g., ‘pvals_entropy_metrics_Age_E16.5.csv’). The CSV must include the following columns: ‘Psi q-value’, ‘Zeta q-value’, ‘Psi’, and ‘Zeta’.

  • save_dir (str, Required) – Directory where the filtered results CSV will be saved.

  • partition_label (Required) – Name of partition used to generate entropy metrics, used to label saved csv.

  • psi_thresh (float, default = 0.5) – Threshold for Psi values. Only genes with Psi > psi_thresh are kept.

  • zeta_thresh (float, Required, default = 0.5) – Threshold for Zeta values. Only genes with Zeta > zeta_thresh are kept.

  • q_thresh (float, default = 0.05) – Threshold for q-values. Genes are retained if both ‘Psi q-value’ <= q_thresh and ‘Zeta q-value’ <= q_thresh.

Returns:

DataFrame containing the significant and specific genes that meet all threshold criteria. Also saved as “highly_specific_genes_to_{partition_label}.csv” in the specified save directory.

Return type:

pd.DataFrame

ember_py.top_genes.non_specific_to_partition(partition_label, pvals_dir, save_dir, psi_thresh=0.5, zeta_thresh=0.5, q_thresh=0.05)

Identifies significant and non-specific genes from a ember generated p-values/q-values CSV file based on thresholds for Psi, Zeta, and q-values. (Potential housekeeping genes)

This function reads a CSV file containing Psi and Zeta metrics (and their corresponding q-values), filters genes that meet given significance and specificity thresholds, and saves the resulting subset to a new CSV file.

Parameters:

  • pvals_dir (str, Required) – Path to the input CSV file (e.g., ‘pvals_entropy_metrics_Age_E16.5.csv’). The CSV must include the following columns: ‘Psi q-value’, ‘Zeta q-value’, ‘Psi’, and ‘Zeta’.

  • save_dir (str, Required) – Directory where the filtered results CSV will be saved.

  • partition_label (Required) – Name of partition used to generate entropy metrics, used to label saved csv.

  • psi_thresh (float, default = 0.5) – Threshold for Psi values. Only genes with Psi > psi_thresh are kept.

  • zeta_thresh (float, default = 0.5) – Threshold for Zeta values. Only genes with Zeta < zeta_thresh are kept.

  • q_thresh (float, default = 0.05) – Threshold for q-values. Genes are retained if both ‘Psi q-value’ <= q_thresh and ‘Zeta q-value’ <= q_thresh.

Returns:

DataFrame containing the significant and specific genes that meet all threshold criteria. Also saved as “non_specific_genes_to_{partition_label}.csv” in the specified save directory.

Return type:

pd.DataFrame

Module contents