dotools_py.pl.ora_network

dotools_py.pl.ora_network(df, term_key, pval_key, score_key, genes_key, color_key=None, shape_key=None, figsize=(6, 5), palette='tab30', title=None, title_fontproperties=None, ax=None, legend_title=None, legend_properties=None, legend_ncols=1, legend_loc='center left', path=None, filename='Network.svg', show=True, shapes=None, clustering_algorithm='hierarchical', cluster_method='complete', cluster_threshold=0.7, cluster_resolution=1, cluster_criterion='distance', min_cluster_size=3, nx_layout=<function spring_layout>, nx_layout_kwargs=None, labels_fontproperties=None, edge_color='gray', edge_alpha=0.75, textwrap_width=25)

Similarity network of overrepresentation analysis results.

Construct a network in which each node represents an enriched term and edges connect terms that share similar gene sets. Term similarity is computed from the overlap of the genes associated with each term and is used to cluster the network and determine the graph layout. Node size is scaled according to the provided score.

Parameters:

df DataFrame

Pandas dataframe with ora results.

term_key str

Column in the dataframe with the term.

pval_key str

Column in the dataframe with the p-values. After clustering the most significant term in each group will be selected as the representative.

score_key str

Column in the dataframe with the score. This is used to determine the node size.

genes_key str

Column in the dataframe with the genes associated to the term. Needs to be a string with genes comma, or semicolon separated.

color_key str | None (default: None)

Column in the dataframe with a first category to be used to determine the node color.

shape_key str | None (default: None)

Column in dataframe with a second category to be used to determine the shape of the nodes.

figsize tuple (default: (6, 5))

Figure size, the format is (width, height).

palette str | dict (default: 'tab30')

String denoting matplotlib colormap. If not set, it will try to access adata.uns[hue_colors | x_axis_colors], if not the colormap do.utility.tab30() will be used. A dictionary with the categories available in adata.obs[x_axis] or adata.obs[hue] if hue is not None can also be provided. The format is {category:color}.

title str | None (default: None)

Title for the figure.

title_fontproperties Optional[Dict[Literal['size', 'weight'], str | int]] (default: None)

Dictionary which should contain ‘size’ and ‘weight’ to define the fontsize and fontweight of the title of the figure.

ax Axes | None (default: None)

Matplotlib axes to use for plotting. If not set, a new figure will be generated.

legend_title str | None (default: None)

Title for the legend.

legend_properties Optional[Dict[Literal['size', 'weight'], str | int]] (default: None)

Dictionary which should contain ‘size’ and ‘weight’ to define the fontsize and fontweight of the title of the legend.

legend_ncols int (default: 1)

Number of columns for the legend.

legend_loc Literal['center left', 'cemter right', 'upper right', 'upper left', 'lower left', 'lower right', 'right', 'lower center', 'upper center', 'center'] (default: 'center left')

Location of the legend.

path str | PathLike[str] | Path | None (default: None)

Path to the folder to save the figure.

filename str (default: 'Network.svg')

Name of file to use when saving the figure.

show bool (default: True)

If set to False, returns a dictionary with the matplotlib axes.

shapes dict | None (default: None)

Dictionary indicating the marker shape to use for each category in shape_key. The key is a category in df[shape_key] and the value is a valid matplotlib marker

clustering_algorithm Literal['hierarchical', 'louvain', 'connected_components'] (default: 'hierarchical')

Algorithm to use to cluster the terms.

cluster_method str (default: 'complete')

Linkage method to use for calculating clusters.

cluster_threshold float (default: 0.7)

depending on the criterion define the threshold to apply when forming flat clusters or specify the max number of clusters.

cluster_criterion str (default: 'distance')

Criterion to form flat clusters using fcluster

cluster_resolution float (default: 1)

Resolution to use when cluster_algorithm is set to louvain.

min_cluster_size int (default: 3)

Keep terms present in clusters of at least this number of terms.

nx_layout Any (default: <function spring_layout at 0x7b35297fbb00>)

A valid networkx layout.

nx_layout_kwargs Optional[Dict] (default: None)

Additional arguments to pass to the network layout.

labels_fontproperties Optional[Dict[Literal['size', 'weight'], str | int]] (default: None)

Fontsize for the representative terms in each cluster.

edge_color str (default: 'gray')

Color of the edges.

edge_alpha float (default: 0.75)

Alpha value for the edges.

textwrap_width int (default: 25)

The maximum width to use when wrapping text.

Returns:

Depending on show, returns the plot if set to True or a dictionary with the axes

Example

Create a network with the results of ORA

import dotools_py as do
table_go = do.dt.example_ora()
table_go = table_go[table_go['Adjusted P-value'] < 0.2]
table_go["Term"] = table_go["Term"].str.split("\(G").str[0]

# Network with only the enriched terms
do.pl.ora_network(
    table_go[table_go["state"] == "enriched"], term_key="Term",
    pval_key="Adjusted P-value", score_key="Combined Score", genes_key="Genes", figsize=(10, 6),
    min_cluster_size=5
)

Create a network and color by a categorical column

table_go = table_go[table_go['Adjusted P-value'] < 0.15]

do.pl.ora_network(
    table_go, term_key="Term", color_key="state", shape_key="state",
    pval_key="Adjusted P-value", score_key="Combined Score", genes_key="Genes", figsize=(10, 6),
    min_cluster_size=2, palette={"enriched":"firebrick", "depleted":"royalblue"},
    clustering_algorithm="louvain", nx_layout_kwargs={"k":0.5}, edge_alpha=0.75,
)  # Keep clusters with at least 2 terms