Move to scanpy 1.9.3 and other improvements

.github/workflows/python-package.yml

@@ -2,54 +2,48 @@ name: Python package
 
 on: [pull_request]
 
+defaults:
+  run:
+    # for conda env activation
+    shell: bash -l {0}
+
 jobs:
   build:
 
     runs-on: ubuntu-latest
     strategy:
       matrix:
-        python-version: [3.7, 3.8]
+        python-version: ["3.8", "3.9"]
 
     steps:
-
-    - uses: actions/checkout@v2
-      with:
-        path: scanpy-scripts
-
-    - uses: psf/black@stable
-      with:
-        options: '--check --verbose --include="\.pyi?$" .'
-
     - uses: actions/checkout@v2
-      with:
-        repository: theislab/scanpy 
-        path: scanpy
-        ref: 1.8.1
-
-    - name: Setup BATS
-      uses: mig4/setup-bats@v1
-      with:
-        bats-version: 1.2.1
 
-    - name: Set up Python ${{ matrix.python-version }}
-      uses: actions/setup-python@v2
+    - name: Setup mamba
+      uses: mamba-org/provision-with-micromamba@main
       with:
-        python-version: ${{ matrix.python-version }}
-
-    - name: Install dependencies
+        environment-file: test-env.yaml
+        cache-downloads: true
+        channels: conda-forge, bioconda, defaults
+        extra-specs: |
+          python=${{ matrix.python-version }}
+
+    - name: Run black manually
       run: |
-        pushd scanpy
-        patch -p1 < ../scanpy-scripts/scrublet.patch
-        popd
+        black --check --verbose ./
 
-        sudo apt-get install libhdf5-dev
-        pip install -U setuptools>=40.1 wheel 'cmake<3.20' pytest
-        pip install $(pwd)/scanpy-scripts
-        python -m pip install $(pwd)/scanpy --no-deps --ignore-installed -vv
+    # - name: Install dependencies
+    #   run: |
+    #     sudo apt-get install libhdf5-dev
+    #     pip install -U setuptools>=40.1 wheel 'cmake<3.20' pytest
+    #     pip install $(pwd)/scanpy-scripts
+    #     # python -m pip install $(pwd)/scanpy --no-deps --ignore-installed -vv
 
     - name: Run unit tests
-      run: pytest --doctest-modules -v ./scanpy-scripts
+      run: |
+        # needed for __version__ to be available
+        pip install . --no-deps --ignore-installed
+        pytest --doctest-modules -v ./
 
     - name: Test with bats
       run: |
-        ./scanpy-scripts/scanpy-scripts-tests.bats
+        ./scanpy-scripts-tests.bats

.gitignore

@@ -7,3 +7,6 @@
 *.pyc
 /.*history
 /.*swp
+data
+compressed
+uncompressed

README.md

@@ -4,12 +4,22 @@ A command-line interface for functions of the Scanpy suite, to facilitate flexib
 
 ## Install
 
+The recommended way of installing scanpy-scripts is via conda:
+
 ```bash
 conda install scanpy-scripts
-# or
-pip3 install scanpy-scripts
 ```
 
+pip installation is also possible, however the version of mnnpy is not patched as in the conda version, and so the `integrate` command will not work.
+
+```bash
+pip install scanpy-scripts
+```
+
+For development installation, we suggest following the github actions python-package.yml file.
+
+Currently, tests run on python 3.9, so those are the recommended versions if not installing via conda. BKNN doesn't currently install on Python 3.10 due to a skip in Bioconda.
+
 ## Test installation
 
 There is an example script included:
@@ -22,7 +32,7 @@ This requires the [bats](https://github.com/sstephenson/bats) testing framework
 
 ## Commands
 
-Available commands are described below. Each has usage instructions available via --help, consult function documentation in scanpy for further details.
+Available commands are described below. Each has usage instructions available via `--help`, consult function documentation in scanpy for further details.
 
 ```
 Usage: scanpy-cli [OPTIONS] COMMAND [ARGS]...

scanpy-scripts-tests.bats

@@ -653,17 +653,18 @@ setup() {
 }
 
 # Do MNN batch correction, using clustering as batch (just for test purposes)
-
-@test "Run MNN batch integration using clustering as batch" {
-    if [ "$resume" = 'true' ] && [ -f "$mnn_obj" ]; then
-        skip "$mnn_obj exists and resume is set to 'true'"
-    fi
-
-    run rm -f $mnn_obj && eval "$scanpy integrate mnn $mnn_opt $louvain_obj $mnn_obj"
-
-    [ "$status" -eq 0 ]
-    [ -f  "$mnn_obj" ]
-}
+# Commented as it fails with scanpy 1.9.1 
+#
+# @test "Run MNN batch integration using clustering as batch" {
+#    if [ "$resume" = 'true' ] && [ -f "$mnn_obj" ]; then
+#        skip "$mnn_obj exists and resume is set to 'true'"
+#    fi
+#
+#    run rm -f $mnn_obj && eval "$scanpy integrate mnn $mnn_opt $louvain_obj $mnn_obj"
+#
+#    [ "$status" -eq 0 ]
+#    [ -f  "$mnn_obj" ]
+#}
 
 # Do ComBat batch correction, using clustering as batch (just for test purposes)
 

scanpy_scripts/__init__.py

@@ -1,9 +1,9 @@
 """
 Provides version, author and exports
 """
-import pkg_resources
+import importlib.metadata
 
-__version__ = pkg_resources.get_distribution("scanpy-scripts").version
+__version__ = importlib.metadata.version("scanpy-scripts")
 
 __author__ = ", ".join(
     [

scanpy_scripts/cmd_utils.py

@@ -6,10 +6,11 @@
 import pandas as pd
 import scanpy as sc
 import scanpy.external as sce
+
 from .cmd_options import CMD_OPTIONS
 from .lib._paga import plot_paga
-from .obj_utils import _save_matrix
 from .lib._scrublet import plot_scrublet
+from .obj_utils import _save_matrix
 
 
 def make_subcmd(cmd_name, func, cmd_desc, arg_desc, opt_set=None):
@@ -313,6 +314,7 @@ def plot_function(
         showfig = True
         if output_fig:
             import os
+
             import matplotlib.pyplot as plt
 
             sc.settings.figdir = os.path.dirname(output_fig) or "."

scanpy_scripts/lib/_diffexp.py

@@ -2,9 +2,11 @@
 scanpy diffexp
 """
 
+import logging
+import math
+
 import pandas as pd
 import scanpy as sc
-import logging
 
 
 def diffexp(
@@ -22,6 +24,15 @@ def diffexp(
 ):
     """
     Wrapper function for sc.tl.rank_genes_groups.
+
+    Test that we can load a single group.
+    >>> import os
+    >>> from pathlib import Path
+    >>> adata = sc.datasets.krumsiek11()
+    >>> tbl = diffexp(adata, groupby='cell_type', groups='Mo', reference='progenitor')
+    >>> # get the size of the data frame
+    >>> tbl.shape
+    (11, 8)
     """
     if adata.raw is None:
         use_raw = False
@@ -51,6 +62,11 @@ def diffexp(
                 "Singlet groups removed before passing to rank_genes_groups()"
             )
 
+    # avoid issue when groups is a single group as a string simplified by click
+    # https://github.com/ebi-gene-expression-group/scanpy-scripts/issues/123
+    if groups != "all" and isinstance(groups, str):
+        groups = [groups]
+
     sc.tl.rank_genes_groups(
         adata,
         use_raw=use_raw,
@@ -64,17 +80,32 @@ def diffexp(
     de_tbl = extract_de_table(adata.uns[diff_key])
 
     if isinstance(filter_params, dict):
+        key_filtered = diff_key + "_filtered"
         sc.tl.filter_rank_genes_groups(
             adata,
             key=diff_key,
-            key_added=diff_key + "_filtered",
+            key_added=key_filtered,
             use_raw=use_raw,
             **filter_params,
         )
 
-        de_tbl = extract_de_table(adata.uns[diff_key + "_filtered"])
+        # there are non strings on recarray object at this point, in
+        # adata.uns['rank_genes_groups_filtered']['names']
+        # for instance:
+        # adata.uns['rank_genes_groups_filtered']['names'][0]
+        # (nan, nan, 'NKG7', nan, nan, 'PPBP')
+        # this now upsets h5py > 3.0
+        de_tbl = extract_de_table(adata.uns[key_filtered])
         de_tbl = de_tbl.loc[de_tbl.genes.astype(str) != "nan", :]
 
+        # change nan for strings in adata.uns['rank_genes_groups_filtered']['names']
+        # TODO on scanpy updates, check if this is not done within scanpy so that we can remove this
+        for row in range(0, len(adata.uns[key_filtered]["names"])):
+            for col in range(0, len(adata.uns[key_filtered]["names"][row])):
+                element = adata.uns[key_filtered]["names"][row][col]
+                if isinstance(element, float) and math.isnan(element):
+                    adata.uns[key_filtered]["names"][row][col] = "nan"
+
     if save:
         de_tbl.to_csv(save, sep="\t", header=True, index=False)
 

scanpy_scripts/lib/_filter.py

@@ -37,7 +37,7 @@ def filter_anndata(
             k_mito = gene_names.str.startswith("MT-")
             if k_mito.sum() > 0:
                 adata.var["mito"] = k_mito
-                adata.var["mito"] = adata.var["mito"].astype("category")
+                # adata.var["mito"] = adata.var["mito"].astype("category")
             else:
                 logging.warning(
                     "No MT genes found, skip calculating "

scanpy_scripts/lib/_louvain.py

@@ -3,6 +3,7 @@
 """
 
 import scanpy as sc
+
 from ..obj_utils import write_obs
 
 

scanpy_scripts/lib/_mnn.py

@@ -2,9 +2,10 @@
 scanpy external mnn
 """
 
-import scanpy.external as sce
-import numpy as np
 import click
+import numpy as np
+import scanpy.external as sce
+import logging
 
 # Wrapper for mnn allowing use of non-standard slot
 
@@ -16,6 +17,10 @@ def mnn_correct(adata, key=None, key_added=None, var_subset=None, layer=None, **
 
     # mnn will use .X, so we need to put other layers there for processing
 
+    logging.warning(
+        "Use mnn_correct at your own risk, environment installation seems faulty for this module."
+    )
+
     if layer:
         adata.layers["X_backup"] = adata.X
         adata.X = adata.layers[layer]

scanpy_scripts/lib/_norm.py

@@ -3,6 +3,7 @@
 """
 
 import scanpy as sc
+import math
 
 
 def normalize(adata, log_transform=True, **kwargs):
@@ -12,6 +13,13 @@ def normalize(adata, log_transform=True, **kwargs):
     """
     sc.pp.normalize_total(adata, **kwargs)
     if log_transform:
-        sc.pp.log1p(adata)
+        # Natural logarithm is the default by scanpy, if base is not set
+        base = math.e
+        sc.pp.log1p(adata, base=base)
+        # scanpy is not setting base in uns['log1p'] keys, but later on asking for it
+        if "log1p" in adata.uns_keys() and "base" not in adata.uns["log1p"]:
+            # Note that setting base to None doesn't solve the problem at other modules that check for base later on
+            # as adata.uns["log1p"]["base"] = None gets dropped at either anndata write or read.
+            adata.uns["log1p"]["base"] = base
 
     return adata

setup.py

@@ -1,11 +1,11 @@
-from setuptools import setup, find_packages
+from setuptools import find_packages, setup
 
 with open("README.md", "r") as fh:
     long_description = fh.read()
 
 setup(
     name="scanpy-scripts",
-    version="1.1.6",
+    version="1.1.9",
     author="nh3",
     author_email="[email protected]",
     description="Scripts for using scanpy from the command line",
@@ -35,23 +35,24 @@
         ]
     ),
     install_requires=[
-        "packaging",
-        "anndata",
-        "scipy",
-        "matplotlib",
-        "pandas",
-        "h5py<3.0.0",
-        "scanpy==1.8.1",
+        # "packaging",
+        # "anndata",
+        # "scipy",
+        # "matplotlib",
+        # "pandas",
+        # "h5py<3.0.0",
+        "scanpy==1.9.3",
         "louvain",
+        "igraph",
         "leidenalg",
         "loompy",
         "Click<8",
-        "umap-learn",
+        # "umap-learn",
         "harmonypy>=0.0.5",
         "bbknn>=1.5.0",
         "mnnpy>=0.1.9.5",
         "scrublet",
-        "scikit-misc",
+        # "scikit-misc",
         "fa2",
     ],
 )