Mercurial > repos > ebi-gxa > decoupler_pseudobulk
changeset 16:508a93e34599 draft
planemo upload for repository https://github.com/ebi-gene-expression-group/container-galaxy-sc-tertiary/ commit 487508282bda9dbb68138d5c7091f46ef54fe52a
| author | ebi-gxa |
|---|---|
| date | Wed, 19 Feb 2025 16:55:58 +0000 |
| parents | 09c833d9b03b |
| children | 2557d7869e78 |
| files | decoupler_pseudobulk.py decoupler_pseudobulk.xml |
| diffstat | 2 files changed, 145 insertions(+), 1 deletions(-) [+] |
line wrap: on
line diff
--- a/decoupler_pseudobulk.py Fri Nov 29 11:34:16 2024 +0000 +++ b/decoupler_pseudobulk.py Wed Feb 19 16:55:58 2025 +0000 @@ -2,6 +2,7 @@ import anndata import decoupler +import numpy as np import pandas as pd @@ -34,6 +35,60 @@ ) +def create_pseudo_replicates(adata, sample_key, num_replicates, seed=None): + """ + Create pseudo replicates for each sample in the sample_key groups. + + Parameters + ---------- + adata : anndata.AnnData + The AnnData object. + sample_key : str + The column in adata.obs that defines the samples. + num_replicates : int + Number of pseudo replicates to create per sample. + + Returns + ------- + anndata.AnnData + The AnnData object with pseudo replicates. + + Examples + -------- + >>> import anndata + >>> import pandas as pd + >>> import numpy as np + >>> data = { + ... 'obs': pd.DataFrame({'sample': ['A', 'A', 'B', 'B']}), + ... 'X': np.array([[1, 0], [0, 1], [1, 1], [0, 0]]) + ... } + >>> adata = anndata.AnnData(X=data['X'], obs=data['obs']) + >>> adata = create_pseudo_replicates(adata, 'sample', 2) + >>> adata.obs['sample_pseudo'].tolist() + ['A_rep1', 'A_rep2', 'B_rep1', 'B_rep2'] + """ + if seed is not None: + np.random.seed(seed) + + new_sample_key = f"{sample_key}_pseudo" + adata.obs[new_sample_key] = adata.obs[sample_key].astype(str) + + for sample in adata.obs[sample_key].unique(): + sample_indices = adata.obs[ + adata.obs[sample_key] == sample].index.to_numpy() + np.random.shuffle(sample_indices) # Shuffle the indices to randomize + replicate_size = int(len(sample_indices) / num_replicates) + for i in range(num_replicates): + start_idx = i * replicate_size + end_idx = start_idx + replicate_size + replicate_indices = sample_indices[start_idx:end_idx] + adata.obs.loc[replicate_indices, new_sample_key] = ( + adata.obs.loc[replicate_indices, new_sample_key] + f"_rep{i+1}" + ) + + return adata + + def prepend_c_to_index(index_value): if index_value and index_value[0].isdigit(): return "C" + index_value @@ -307,6 +362,13 @@ factor_fields = args.factor_fields.split(",") check_fields(factor_fields, adata) + # Create pseudo replicates if specified + if args.num_pseudo_replicates: + adata = create_pseudo_replicates( + adata, args.sample_key, args.num_pseudo_replicates, seed=args.seed + ) + args.sample_key = f"{args.sample_key}_pseudo" + print(f"Using mode: {args.mode}") # Perform pseudobulk analysis pseudobulk_data = get_pseudobulk( @@ -664,6 +726,19 @@ help="Minimum total count threshold for filtering by expression", ) parser.add_argument( + "--num_pseudo_replicates", + type=int, + choices=range(3, 1000), + help="Number of pseudo replicates to create per sample (at least 3)", + required=False + ) + parser.add_argument( + "--seed", + type=int, + default=None, + help="Random seed for pseudo replicate sampling", + ) + parser.add_argument( "--anndata_output_path", type=str, help="Path to save the filtered AnnData object or pseudobulk data",
--- a/decoupler_pseudobulk.xml Fri Nov 29 11:34:16 2024 +0000 +++ b/decoupler_pseudobulk.xml Wed Feb 19 16:55:58 2025 +0000 @@ -1,4 +1,4 @@ -<tool id="decoupler_pseudobulk" name="Decoupler pseudo-bulk" version="1.4.0+galaxy8" profile="20.05"> +<tool id="decoupler_pseudobulk" name="Decoupler pseudo-bulk" version="1.4.0+galaxy9" profile="20.05"> <description>aggregates single cell RNA-seq data for running bulk RNA-seq methods</description> <requirements> <requirement type="package" version="1.4.0">decoupler</requirement> @@ -47,6 +47,12 @@ --contrasts_file '$filter_per_contrast.contrasts_file' --min_gene_exp_perc_per_cell '$filter_per_contrast.min_cells_perc_per_contrast_cond' #end if + #if $num_pseudo_replicates: + --num_pseudo_replicates $num_pseudo_replicates + #if $seed: + --seed '$seed' + #end if + #end if --deseq2_output_path deseq_output_dir --plot_samples_figsize $plot_samples_figsize --plot_filtering_figsize $plot_filtering_figsize @@ -89,6 +95,8 @@ <param type="boolean" name="filter_expr" label="Enable Filtering by Expression"/> <param type="text" name="plot_samples_figsize" label="Plot Samples Figsize" value="10 10" help="X and Y sizes in points separated by a space"/> <param type="text" name="plot_filtering_figsize" label="Plot Filtering Figsize" value="10 10" help="X and Y sizes in points separated by a space"/> + <param type="integer" name="num_pseudo_replicates" label="Number of Pseudo Replicates" optional="true" help="If set, create this number of pseudo replicates to create per sample (at least 3). If not set, there is an expectation that samples and groups are distributed in a way that (pseudo) replicates exists." min="3" max="1000"/> + <param type="integer" name="seed" label="Seed" optional="true" help="Seed for the random number generator used for sampling the pseudo replicates (only used if Number of Pseudo replicates set)."/> </inputs> <outputs> <data name="pbulk_anndata" format="h5ad" label="${tool.name} on ${on_string}: Pseudo-bulk AnnData"> @@ -229,6 +237,66 @@ </assert_contents> </output> </test> + <test expect_num_outputs="8"> + <param name="input_file" value="mito_counted_anndata.h5ad"/> + <param name="filter" value="yes"/> + <param name="contrasts_file" value="test_contrasts.txt" ftype="txt"/> + <param name="min_cells_perc_per_contrast_cond" value="25"/> + <param name="adata_obs_fields_to_merge" value="batch,sex:batch,genotype"/> + <param name="groupby" value="batch_sex"/> + <param name="sample_key" value="genotype"/> + <param name="factor_fields" value="genotype,batch_sex"/> + <param name="mode" value="sum"/> + <param name="min_cells" value="10"/> + <param name="produce_plots" value="true"/> + <param name="produce_anndata" value="true"/> + <param name="min_counts" value="10"/> + <param name="min_counts_per_sample" value="50"/> + <param name="min_total_counts" value="1000"/> + <param name="filter_expr" value="true"/> + <param name="plot_samples_figsize" value="10 10"/> + <param name="plot_filtering_figsize" value="10 10"/> + <param name="num_pseudo_replicates" value="3"/> + <param name="seed" value="42"/> + <output name="pbulk_anndata" ftype="h5ad"> + <assert_contents> + <has_h5_keys keys="obs/psbulk_n_cells"/> + </assert_contents> + </output> + <output name="count_matrix" ftype="tabular"> + <assert_contents> + <has_n_lines n="3620"/> + <has_n_columns n="22"/> + </assert_contents> + </output> + <output name="samples_metadata" ftype="tabular"> + <assert_contents> + <has_n_lines n="22"/> + <has_n_columns n="3"/> + </assert_contents> + </output> + <output name="genes_metadata" ftype="tabular"> + <assert_contents> + <has_n_lines n="3620"/> + <has_n_columns n="13"/> + </assert_contents> + </output> + <output name="plot_output" ftype="png"> + <assert_contents> + <has_size value="34626" delta="6000"/> + </assert_contents> + </output> + <output name="filter_by_expr_plot" ftype="png"> + <assert_contents> + <has_size value="21656" delta="2000"/> + </assert_contents> + </output> + <output name="genes_ignore_per_contrast" ftype="tabular"> + <assert_contents> + <has_n_lines n="35478"/> + </assert_contents> + </output> + </test> </tests> <help> <![CDATA[ @@ -250,6 +318,7 @@ - Enable Filtering by Expression: Check this box to enable filtering by expression. - Plot Samples Figsize: Size of the samples plot as a tuple (two arguments). - Plot Filtering Figsize: Size of the filtering plot as a tuple (two arguments). + - Number of Pseudo Replicates: Number of pseudo replicates to create per sample (at least 3). The tool will output the filtered AnnData, count matrix, samples metadata, genes metadata (in DESeq2 format), and the pseudobulk plot and filter by expression plot (if enabled). Files for filtering genes later on are also generated (to ignore after the DE model).