added documentation

dingo/illustrations.py

@@ -135,6 +135,18 @@ def plot_corr_matrix(corr_matrix, reactions, removed_reactions=[], format="svg")
 
 
 def plot_dendrogram(dissimilarity_matrix, reactions , plot_labels=False, t=2.0, linkage="ward"):
+ """A Python function to plot the dendrogram of a dissimilarity matrix.
+
+ Keyword arguments:
+ dissimilarity_matrix -- A matrix produced from the "cluster_corr_reactions" function
+ reactions -- A list with the model's reactions
+ plot_labels -- A boolean variable that if True plots the reactions labels in the dendrogram
+ t -- A threshold that defines a threshold that cuts the dendrogram 
+ at a specific height and colors the occuring clusters accordingly
+ linkage -- linkage defines the type of linkage. 
+ Available linkage types are: single, average, complete, ward.
+ """
+
  fig = ff.create_dendrogram(dissimilarity_matrix,
  labels=reactions,
  linkagefun=lambda x: hierarchy.linkage(x, linkage), 
@@ -160,6 +172,13 @@ def plot_dendrogram(dissimilarity_matrix, reactions , plot_labels=False, t=2.0,
 
 
 def plot_graph(G, pos):
+ """A Python function to plot a graph created from a correlation matrix.
+
+ Keyword arguments:
+ G -- A graph produced from the "graph_corr_matrix" function.
+ pos -- A layout for the corresponding graph.
+ """
+
  fig = go.Figure()
 
  for u, v, data in G.edges(data=True):
@@ -182,7 +201,7 @@ def plot_graph(G, pos):
  text=[node_name],
  textposition='top center',
  name = node_name,
- showlegend=True))
+ showlegend=False))
 
  fig.update_layout(width=800, height=800)
  fig.show()

dingo/utils.py

@@ -346,10 +346,21 @@ def correlated_reactions(steady_states, reactions=[], pearson_cutoff = 0.90, ind
 
 
 
-def cluster_corr_reactions(correlation_matrix, reactions, method, 
- linkage="ward", t = 4.0,
- min_cluster_size = 2, n_clusters=8):
-
+def cluster_corr_reactions(correlation_matrix, reactions, linkage="ward", 
+ t = 4.0, correction=True):
+ """A Python function for hierarchical clustering of the correlation matrix
+
+ Keyword arguments:
+ correlation_matrix -- A numpy 2D array of a correlation matrix
+ reactions -- A list with the model's reactions
+ linkage -- linkage defines the type of linkage. 
+ Available linkage types are: single, average, complete, ward.
+ t -- A threshold that defines a threshold that cuts the dendrogram 
+ at a specific height and produces clusters
+ correction -- A boolean variable that if True converts the values of the
+ the correlation matrix to absolute values.
+ """
+
  # function to return a nested list with grouped reactions based on clustering
  def clusters_list(reactions, labels):
  clusters = []
@@ -362,29 +373,39 @@ def clusters_list(reactions, labels):
  clusters.append(cluster)
  return clusters
 
- if method == "hierarchical":
- dissimilarity_matrix = 1 - abs(correlation_matrix) 
-  Z = hierarchy.linkage(dissimilarity_matrix, linkage)
- labels = hierarchy.fcluster(Z, t, criterion='distance')
-
-  clusters = clusters_list(reactions, labels)
-  return dissimilarity_matrix, labels, clusters
+ if correction == True: 
+ dissimilarity_matrix = 1 - abs(correlation_matrix)
+ else:
+ dissimilarity_matrix = 1 - correlation_matrix
+  
+ Z = hierarchy.linkage(dissimilarity_matrix, linkage)
+ labels = hierarchy.fcluster(Z, t, criterion='distance')
 
+ clusters = clusters_list(reactions, labels)
+ return dissimilarity_matrix, labels, clusters
 
+
+
+def graph_corr_matrix(correlation_matrix, reactions, correction=True, 
+ clusters=[], subgraph_nodes = 5):
+ """A Python function that creates the main graph and its subgraphs
+ from a correlation matrix.
 
-def graph_corr_matrix(correlation_matrix, reactions, correction=True, clusters=[]):
+ Keyword arguments:
+ correlation_matrix -- A numpy 2D array of a correlation matrix.
+ reactions -- A list with the model's reactions.
+ correction -- A boolean variable that if True converts the values of the
+ the correlation matrix to absolute values.
+ clusters -- A nested list with clustered reactions created from the "" function.
+ subgraph_nodes -- A variable that specifies a cutoff for a graph's nodes. 
+ It filters subgraphs with low number of nodes..
+ """
 
  graph_matrix = correlation_matrix.copy()
  np.fill_diagonal(graph_matrix, 0)
 
  if correction == True:
- for cluster in clusters:
- for reaction_a in cluster:
- for reaction_b in cluster:
- reaction_a_index = reactions.index(reaction_a)
- reaction_b_index = reactions.index(reaction_b)
- graph_matrix[reaction_a_index, reaction_b_index] = abs(graph_matrix[reaction_a_index, reaction_b_index])
- graph_matrix[reaction_b_index, reaction_a_index] = abs(graph_matrix[reaction_b_index, reaction_a_index])
+ graph_matrix = abs(graph_matrix)
 
  G = nx.from_numpy_array(graph_matrix)
  G = nx.relabel_nodes(G, lambda x: reactions[x])
@@ -393,7 +414,6 @@ def graph_corr_matrix(correlation_matrix, reactions, correction=True, clusters=[
  unconnected_nodes = list(nx.isolates(G))
  G.remove_nodes_from(unconnected_nodes)
  G_nodes = G.nodes()
- print(G_nodes)
 
  graph_list = []
  layout_list = []
@@ -405,11 +425,12 @@ def graph_corr_matrix(correlation_matrix, reactions, correction=True, clusters=[
  H_nodes_list = []
 
  for i in range(len(subgraphs)):
- if len(subgraphs[i].nodes()) > 4 and len(subgraphs[i].nodes()) != len(G_nodes): 
+ if len(subgraphs[i].nodes()) > subgraph_nodes and len(subgraphs[i].nodes()) != len(G_nodes): 
  H = G.subgraph(subgraphs[i].nodes())
  for cluster in clusters:
  if H.has_node(cluster[0]) and H.nodes() not in H_nodes_list:
  H_nodes_list.append(H.nodes())
+
  pos = nx.spring_layout(H)
  graph_list.append(H)
  layout_list.append(pos)