46 expand tests

networkcommons/_utils.py

@@ -37,26 +37,31 @@ def edge_attrs_from_corneto(graph: cn.Graph) -> pd.DataFrame:
     concat_df.rename(columns={0: 'node'}, inplace=True)
 
 
-
 def to_cornetograph(graph):
     """
     Convert a networkx graph to a corneto graph, if needed.
 
     Args:
-        graph (nx.Graph or nx.DiGraph): The corneto graph.
+        graph (nx.DiGraph): The corneto graph.
 
     Returns:
         cn.Graph: The corneto graph.
     """
-    if isinstance(graph, cn._graph.Graph):
+    if isinstance(graph, nx.MultiDiGraph):
+        raise NotImplementedError("Only nx.DiGraph graphs and corneto graphs are supported.")
+    elif isinstance(graph, cn.Graph):
         corneto_graph = graph
-    elif isinstance(graph, (nx.Graph, nx.DiGraph)):
+    elif isinstance(graph, nx.DiGraph):
         # substitute 'sign' for 'interaction' in the graph
         nx_graph = graph.copy()
         for u, v, data in nx_graph.edges(data=True):
             data['interaction'] = data.pop('sign')
 
         corneto_graph = cn_nx.networkx_to_corneto_graph(nx_graph)
+    elif isinstance(graph, nx.Graph):
+        raise NotImplementedError("Only nx.DiGraph graphs and corneto graphs are supported.")
+    else:
+        raise NotImplementedError("Only nx.DiGraph graphs and corneto graphs are supported.")
 
     return corneto_graph
 
@@ -71,15 +76,21 @@ def to_networkx(graph, skip_unsupported_edges=True):
     Returns:
         nx.Graph: The networkx graph.
     """
-    if isinstance(graph, nx.Graph) or isinstance(graph, nx.DiGraph):
+    if isinstance(graph, nx.MultiDiGraph):
+        raise NotImplementedError("Only nx.DiGraph graphs and corneto graphs are supported.")
+    elif isinstance(graph, nx.DiGraph):
         networkx_graph = graph
-    elif isinstance(graph, cn._graph.Graph):
+    elif isinstance(graph, cn.Graph):
         networkx_graph = cn_nx.corneto_graph_to_networkx(
             graph,
             skip_unsupported_edges=skip_unsupported_edges)
         # rename interaction for sign
         for u, v, data in networkx_graph.edges(data=True):
             data['sign'] = data.pop('interaction')
+    elif isinstance(graph, nx.Graph):
+        raise NotImplementedError("Only nx.DiGraph graphs and corneto graphs are supported.")
+    else:
+        raise NotImplementedError("Only nx.DiGraph graphs and corneto graphs are supported.")
 
     return networkx_graph
 
@@ -116,8 +127,7 @@ def read_network_from_file(file_path,
 def network_from_df(network_df,
                     source_col='source',
                     target_col='target',
-                    directed=True,
-                    multigraph=False):
+                    directed=True):
     """
     Create a network from a DataFrame.
 
@@ -132,9 +142,6 @@ def network_from_df(network_df,
     """
     network_type = nx.DiGraph if directed else nx.Graph
 
-    if multigraph:
-        network_type = nx.MultiDiGraph if directed else nx.MultiGraph
-
     if list(network_df.columns) == list([source_col, target_col]):
         network = nx.from_pandas_edgelist(network_df,
                                           source=source_col,
@@ -183,8 +190,8 @@ def decoupler_formatter(df,
     Format dataframe to be used by decoupler.
 
     Parameters:
-        df (DataFrame): A pandas DataFrame.
-        column (str): The column to be used as index.
+        df (DataFrame): A pandas DataFrame. Index should be populated
+        column (str): The columns to be subsetted.
 
     Returns:
         A formatted DataFrame.
@@ -211,7 +218,7 @@ def targetlayer_formatter(df, n_elements=25):
 
     # Sort the DataFrame by the absolute value of the
     # 'sign' column and get top n elements
-    df = df.sort_values(by='sign', key=lambda x: abs(x))
+    df = df.sort_values(by='sign', key=lambda x: abs(x), ascending=False)
 
     df = df.head(n_elements)
 

networkcommons/data/network/_liana.py

@@ -19,9 +19,11 @@
 
 __all__ = ['get_lianaplus']
 
-import lazy_import
+# import lazy_import
 
-liana = lazy_import.lazy_module('liana')
+# liana = lazy_import.lazy_module('liana')
+
+import liana
 
 
 def get_lianaplus(resource='Consensus'):
@@ -37,8 +39,6 @@ def get_lianaplus(resource='Consensus'):
         pandas.DataFrame: Liana+ network with source, target, and sign columns.
     """
 
-    import liana
-
     network = liana.resource.select_resource(resource).drop_duplicates()
     network.columns = ['source', 'target']
     network['sign'] = 1

networkcommons/data/network/_moon.py

@@ -17,7 +17,7 @@
 Prior knowledge network used by MOON.
 """
 
-__all__ = ['build_moon_regulons', 'get_cosmos_pkn']
+__all__ = ['get_cosmos_pkn']
 
 import lazy_import
 import numpy as np
@@ -59,52 +59,3 @@ def get_cosmos_pkn(update: bool = False):
         file_legend = pd.read_pickle(path)
 
     return file_legend
-
-
-
-
-def build_moon_regulons(include_liana=False):
-
-    dorothea_df = dc.get_collectri()
-
-    TFs = np.unique(dorothea_df['source'])
-
-    full_pkn = _omnipath.get_omnipath(genesymbols=True, directed_signed=True)
-
-    if include_liana:
-
-        ligrec_resource = _liana.get_lianaplus()
-
-        full_pkn = pd.concat([full_pkn, ligrec_resource])
-        full_pkn['edgeID'] = full_pkn['source'] + '_' + full_pkn['target']
-
-        # This prioritises edges coming from OP
-        full_pkn = full_pkn.drop_duplicates(subset='edgeID')
-        full_pkn = full_pkn.drop(columns='edgeID')
-
-    kinTF_regulons = full_pkn[full_pkn['target'].isin(TFs)].copy()
-    kinTF_regulons.columns = ['source', 'target', 'mor']
-    kinTF_regulons = kinTF_regulons.drop_duplicates()
-
-    kinTF_regulons = kinTF_regulons.groupby(['source', 'target']).mean() \
-        .reset_index()
-
-    layer_2 = {}
-    activation_pkn = full_pkn[full_pkn['sign'] == 1].copy()
-
-    pkn_graph = _utils.network_from_df(activation_pkn, directed=True)
-
-    relevant_nodes = list(activation_pkn['source'].unique())
-    relevant_nodes = [node for node in relevant_nodes if node in list(kinTF_regulons['source'])]
-
-    for node in relevant_nodes:
-        intermediates = activation_pkn[activation_pkn['source'] == node]['target'].tolist()
-        targets = [n for i in intermediates for n in pkn_graph.neighbors(i)]
-        targets = np.unique([n for n in targets if n in TFs])
-        layer_2[node] = targets
-
-    layer_2_df = pd.concat([pd.DataFrame({'source': k, 'target': v, 'mor': 0.25}) for k, v in layer_2.items()], ignore_index=True)
-    kinTF_regulons = pd.concat([kinTF_regulons, layer_2_df])
-    kinTF_regulons = kinTF_regulons.groupby(['source', 'target']).sum().reset_index()
-
-    return kinTF_regulons

networkcommons/data/omics/__init__.py

@@ -22,5 +22,5 @@
 from ._deseq2 import *
 from ._panacea import *
 from ._scperturb import *
-from ._moon import *
+from ._nci60 import *
 from ._cptac import *

networkcommons/data/omics/_common.py

@@ -75,11 +75,6 @@ def _commons_url(dataset: str, **kwargs) -> str:
     return urllib.parse.urljoin(baseurl, path)
 
 
-def _dataset(key: str) -> dict | None:
-
-    return _datasets()['datasets'].get(key.lower(), None)
-
-
 def _requests_session() -> requests.Session:
 
     ses = requests.Session()

networkcommons/data/omics/_cptac.py

@@ -160,5 +160,6 @@ def cptac_extend_dataframe(df):
 
     extended_df.drop(['Tumor', 'Normal'], axis=1, inplace=True)
     extended_df.rename(columns={'idx': 'sample_ID'}, inplace=True)
+    extended_df.reset_index(inplace=True, drop=True)
 
     return extended_df

networkcommons/data/omics/_deseq2.py

@@ -21,18 +21,13 @@
 
 __all__ = ['deseq2']
 
-from typing import TYPE_CHECKING
 import multiprocessing
-import importlib
 
-if TYPE_CHECKING:
+import pandas as pd
 
-    import pandas as pd
-
-import lazy_import
 from pypath_common import _misc as _ppcommon
 
-#for _mod in ('default_inference', 'dds', 'ds'):
+# for _mod in ('default_inference', 'dds', 'ds'):
 
 #    globals()[f'_deseq2_{_mod}'] = lazy_import.lazy_module(f'pydeseq2.{_mod}')
 
@@ -50,8 +45,7 @@ def deseq2(
         ref_group: str,
         sample_col: str = 'sample_ID',
         feature_col: str = 'gene_symbol',
-        covariates: list | None = None,
-        round_values: bool = False
+        covariates: list | None = None
     ) -> pd.DataFrame:
     """
     Runs DESeq2 analysis on the given counts and metadata.
@@ -67,8 +61,6 @@ def deseq2(
             Defaults to 'gene_symbol'.
         covariates (list, optional): List of covariates to include in the analysis.
             Defaults to None.
-        round_values (bool, optional): Whether to round the counts to integers. Otherwise, the
-            counts will be left as floats and the function will fail. Defaults to False.
 
 
     Returns:
@@ -86,11 +78,6 @@ def deseq2(
     if '_' in ref_group:
         ref_group = ref_group.replace('_', '-')
 
-    if round_values and not counts.select_dtypes(include=['float64', 'float32']).empty:
-        counts = counts.round(0)
-        _log('Float values found. Rounded counts to integers.')
-
-
     n_cpus = _conf.get('cpu_count', multiprocessing.cpu_count())
     inference = _deseq2_default_inference.DefaultInference(n_cpus = n_cpus)
 

networkcommons/data/omics/_scperturb.py

@@ -24,7 +24,6 @@
 from typing import Any
 import json
 
-import bs4
 import anndata as ad
 
 from . import _common

networkcommons/eval/_metrics.py

@@ -189,7 +189,7 @@ def get_graph_metrics(network, target_dict):
 
         metrics.reset_index(inplace=True, drop=True)
 
-    elif isinstance(network, (nx.Graph, nx.DiGraph)):
+    elif isinstance(network, nx.DiGraph):
         metrics = pd.DataFrame({
             'Number of nodes': get_number_nodes(network),
             'Number of edges': get_number_edges(network),
@@ -198,6 +198,8 @@ def get_graph_metrics(network, target_dict):
             'Mean closeness': get_mean_closeness(network),
             'Connected targets': get_connected_targets(network, target_dict)
         }, index=[0])
+    else:
+        raise TypeError("The network must be a networkx graph or a dictionary of networkx graphs.")
 
     return metrics