last version still buggy

debruijn/debruijn.py

@@ -45,7 +45,7 @@ def build_kmer_dict(fichier, k):
         hash_table, dict: dictionnary with key = k-mer as str
                           and value count of k-mer occurence
     """
-    hash_table = {} # initialise empty hash table
+    hash_table = {}# initialise empty hash table
     it_fastq = read_fastq(fichier)
     for seq in it_fastq: # for each sequence
         it_kmer = cut_kmer(seq, k) # count each occurence of k-mer
@@ -132,11 +132,11 @@ def path_average_weight(graph, path):
     """
     weight = 0
     for i in range(len(path)-1):
-        weight += graph[path[i][i+i][weight]
+        weight += graph.edges[path[i], path[i+1]]["weight"]
-    
-    return weight/(len(path)-1)
 
+    weight = weight/(len(path) - 1)
 
+    return weight
 
 def remove_paths(graph, paths, delete_entry_node=False, delete_sink_node=False):
     """
@@ -164,16 +164,30 @@ def remove_paths(graph, paths, delete_entry_node=False, delete_sink_node=False):
     return graph
 
 
-def select_best_path(graph, paths, path_len, mean_weights,
+def select_best_path(graph, paths, path_lens, mean_weights,
                      delete_entry_node=False, delete_sink_node=False):
     """
+    Given path list, their length and weight, keeps only the best path among
+    them considering the following priority : weight, length and randomly
 
+    Arguments:
+        graph, nx.DiGraph: a de bruijn graph
+        paths, list of str: list of paths
+        path_lens: lengths of the paths
+        mean_weights: mean weights of the paths
+        delete_entry_node, boolean: either or not if the entry node
+                                    should be deleted
+        delete_sink_node, boolean: either or not if the sink node
+                                   should be deleted
+
+    Returns:
+        graph, nx.DiGraph: graph with deleted paths
     """
     max_weight = max(mean_weights)
-    heaviest = [i for i, j in enumerate(mean_weights) if j == mean_weights]
+    heaviest = [i for i, j in enumerate(mean_weights) if j == max_weight]
     if len(heaviest) > 1:
-        max_len = max(path_lengths)
+        max_len = max(path_lens)
-        longest = [i for i in heaviest if path_len[i] == max_len]
+        longest = [i for i in heaviest if path_lens[i] == max_len]
         if len(longest) > 1:
             Random.seed(9001)
             best = random.choice[longest]
@@ -181,9 +195,14 @@ def select_best_path(graph, paths, path_len, mean_weights,
             best = longest[0]
     else:
         best = heaviest[0]
-    paths.pop(best)
     
-    return remove_paths(graph, paths, delete_entry_node, delete_sink_node)
+    for p in paths:
+        print(p)
+
+    paths2 = [p for p in paths]
+    paths2.pop(best)
+
+    return remove_paths(graph, paths2, delete_entry_node, delete_sink_node)
 
 
 def fill(text, width=80):
@@ -227,13 +246,47 @@ def get_contigs(graph, starting_nodes, sink_nodes):
     
     return contigs
 
-def solve_bubble():
+def solve_bubble(graph, ancestor_node, descendent_node):
-    pass
+    """
+    solve a bubble
+
+    Arguments:
+        graph, nx.DiGraph: a de bruijn graph
+        ancestor_node, str: a node
+        descendent_node, str: a node
+
+    Returns:
+        graph, nx.DiGraph: the same graph without the bubble
+    """
+    paths = algorithms.all_simple_paths(graph, ancestor_node, descendent_node)
+
+    weights = []
+    path_lens = []
+    for path in paths:# constituting weights and length lists
+        weights.append(path_average_weight(graph, path))
+        path_lens.append(len(path))
+
+    return select_best_path(graph, paths, weights, path_lens) # keep best path
 
 
-def simplify_bubbles():
+def simplify_bubbles(graph):
-    pass
+    """
+    Returns a bubble-less graph 
 
+    Arguments:
+        graph, nx.DiGraph: a de bruijn graph
+
+    Returns:
+        graph, nx.DiGraph: a bubble-less de bruijn graph
+    """
+    fork_nodes = []# empty list containing nodes with 2 or more ancestors
+    for node in graph:
+        while graph.in_degree(node) >= 2: # if 2 or more ancestor add node
+            pred = [n for n in graph.predecessors(node)] 
+            ancestor = algorithms.lowest_common_ancestor(graph,pred[0], pred[1])
+            graph = solve_bubble(graph, ancestor, node)
+
+    return graph
 
 def solve_entry_tips():
     pass