Great question! When you need to treat specific amino acid nodes as equivalent (like LEU and ILE) but not others, NetworkX's subgraph_is_isomorphic() lets you pass a custom node_match function to override the default exact attribute matching. Here's how to implement this for your use case:

Step 1: Define the Custom Node Match Function #

The node_match function takes two node attribute dictionaries (one from each graph) and returns True if the nodes should be considered equivalent. For your scenario, we want to:

• Return True if the labels are identical (e.g., ARG ↔ ARG)
• Return True if one label is LEU and the other is ILE
• Return False for all other pairs

def amino_acid_node_match(n1_attrs, n2_attrs):
    label1 = n1_attrs['label']
    label2 = n2_attrs['label']
    
    # Exact match case
    if label1 == label2:
        return True
    # LEU and ILE are equivalent
    elif (label1 == 'LEU' and label2 == 'ILE') or (label1 == 'ILE' and label2 == 'LEU'):
        return True
    # All other pairs are not equivalent
    return False

Step 2: Use the Function in Isomorphism Check #

Pass this custom function to subgraph_is_isomorphic() to get the desired behavior. Using your example graphs:

import networkx as nx

# Your original graphs
G1 = nx.Graph()
G1.add_node(1, label='ARG')
G1.add_node(2, label='LEU')
G1.add_edge(1, 2)

G2 = nx.Graph()
G2.add_node(1, label='ARG')
G2.add_node(2, label='ILE')
G2.add_edge(1, 2)

# With custom node matching: returns True
is_isomorphic_custom = nx.subgraph_is_isomorphic(G1, G2, node_match=amino_acid_node_match)
print(is_isomorphic_custom)  # Output: True

# Without custom matching: returns False (default exact match)
is_isomorphic_default = nx.subgraph_is_isomorphic(G1, G2)
print(is_isomorphic_default)  # Output: False

Bonus: Scaling to Multiple Equivalent Pairs #

If you need to add more equivalent amino acid pairs later (e.g., VAL ↔ ALA), you can refactor the function to use groups of equivalent labels for easier maintenance:

# Define groups of equivalent amino acids
equivalent_amino_acids = [{'LEU', 'ILE'}, {'VAL', 'ALA'}, {'ASP', 'GLU'}]

def scalable_node_match(n1_attrs, n2_attrs):
    label1 = n1_attrs['label']
    label2 = n2_attrs['label']
    
    if label1 == label2:
        return True
    # Check if both labels belong to the same equivalent group
    for group in equivalent_amino_acids:
        if label1 in group and label2 in group:
            return True
    return False

This approach keeps your code clean as you expand the set of equivalent nodes.

内容的提问来源于stack exchange，提问作者Mahdi Pourbaghi