first commit

Author: ktroule

.gitignore

@@ -0,0 +1 @@
+*.pyc

README.md

@@ -0,0 +1,3 @@
+# sctools
+
+Repository with utilities that I've create for the analysis of single-cell data.

__init__.py

@@ -0,0 +1,74 @@
+from joblib import Parallel, delayed
+import multiprocessing
+import scanpy as sc
+import pandas as pd
+import glob
+import os
+
+def soupy_ratio(data_dir, num_cores = -1):
+    
+    '''
+    Calculates the ratio of gene counts associated to empty droplets to all droplets.
+    
+    This aims to identify soupy genes, those whose expression ratio is high in the empty droplets.
+    A higher score (i.e close to 1), the more likely the gene is soupy.
+    
+    Parameters:
+        - data_dir (int): Folder containing all the libraries associated to a given experiment.
+        - num_cores (int): Number of cores to use, default all unless specified.
+            
+    Parameters:
+        - a dataframe with the ratio of soupiness associated to every gene.
+    
+    '''
+    
+    # -- List all subfolders containing raw and filtered matrices
+    dirs_dropl = glob.glob(os.path.join(data_dir, '*', 'raw_feature_bc_matrix'))
+    dirs_cells = glob.glob(os.path.join(data_dir, '*', 'filtered_feature_bc_matrix'))
+
+    # -- Function to load adata and modify cell barcodes
+    def load_adata(obj_dir):
+        adata = sc.read_10x_mtx(obj_dir)
+        adata.obs.index = os.path.basename(obj_dir.split('/')[-2]) + '_' + adata.obs.index
+        return adata
+
+
+    # -- Set number of cores for parallelization
+    if num_cores <= 0: 
+        num_cores = multiprocessing.cpu_count()
+
+    # -- Read all matrices and concateante into a single object for raw and filtered
+    if __name__ == "__main__":
+        print('Reading raw_feature_bc_matrix objects')
+        dropl_holder = Parallel(n_jobs = num_cores)(delayed(load_adata)(obj_dir = i)
+                                                    for i in dirs_dropl)
+
+        print('Reading filtered_feature_bc_matrix objects')
+        cells_holder = Parallel(n_jobs = num_cores)(delayed(load_adata)(obj_dir = i)
+                                                    for i in dirs_cells)
+
+    # -- Concatenate objets
+    adata_dropl = dropl_holder[0].concatenate(dropl_holder[1:], join = 'outer')
+    del dropl_holder
+    
+    adata_cells = cells_holder[0].concatenate(cells_holder[1:], join = 'outer')
+    del cells_holder
+    
+    # -- From droplets remove barcodes associtate to cells
+    adata_dropl = adata_dropl[~adata_dropl.obs.index.isin(list(adata_cells.obs.index))]
+
+    # -- Calculate number of gene counts 
+    counts_dropl = adata_dropl.X.sum(axis = 0).A1
+    counts_cells = adata_cells.X.sum(axis = 0).A1
+
+    # -- Join counts in a dataframe and discard genes whose counts is 0 in both objects
+    soupy_ratio = pd.DataFrame({'droplets' : counts_dropl, 'cells' : counts_cells},
+                               index = adata_dropl.var.index)
+    
+    # -- Remove genes 0 counts for droplets and cells
+    soupy_ratio = soupy_ratio[soupy_ratio[['droplets', 'cells']].sum(axis = 1) != 0]
+    
+    # -- Calculate ratio of soupiness
+    soupy_ratio['soupy_ratio'] = soupy_ratio['droplets']/soupy_ratio[['droplets', 'cells']].sum(axis = 1)
+
+    return(soupy_ratio)

metrics.py

@@ -0,0 +1,18 @@
+[project]
+name = "sctools"
+version = "0.0.1"
+authors = [
+  { name="Kevin", email="" },
+]
+description = "Utilities for single-cell analyses."
+readme = "README.md"
+requires-python = ">=3.7"
+classifiers = [
+    "Programming Language :: Python :: 3",
+    "License :: OSI Approved :: MIT License",
+    "Operating System :: OS Independent",
+]
+
+[project.urls]
+"Homepage" = ""
+"Bug Tracker" = "https://github.com/ktroule/sctools"