Impact of 'nBits' on GetHashedMorganFingerprint()

Cheminformatics Tutorials - Herong's Tutorial Examples

∟Impact of 'nBits' on GetHashedMorganFingerprint()

This section provides a tutorial example on impact of the 'nBits' option on fingerprint generation with rdkit.Chem.rdMolDescriptors.GetHashedMorganFingerprint() function.

The 'nBits' option in the rdkit.Chem.rdMolDescriptors.GetHashedMorganFingerprint() function call allows you to control the range of smaller integers of the fingerprint. The default is nBits=2048, which is capable to represent 2048 identifiers in a perfect scenario.

Hashing identifiers to smaller integers may create collision conditions where different identifiers are hashed to the same integer.

1. For example, "OOOOOOOO" has 5 different identifiers with radius=1. They will be hashed into 4 different smaller integers with nBits=16.

from rdkit.Chem import AllChem
from rdkit.DataStructs.cDataStructs import UIntSparseIntVect
mol = AllChem.MolFromSmiles('OOOOOOOO')
fp = AllChem.GetMorganFingerprint(mol, 1)
display(UIntSparseIntVect.GetNonzeroElements(fp))

fp = AllChem.GetHashedMorganFingerprint(mol, 1, nBits=16,
  bitInfo=bitInfo)
display(UIntSparseIntVect.GetNonzeroElements(fp))

# output 
{864662311: 2, 864674487: 6, 2379014453: 2, 3705148550: 4, 3706500114: 2}
{2: 2, 5: 2, 6: 4, 7: 8}

Identifier 864662311 and identifier 864674487 are collided, when both are hashed to 7. But their number of duplications 2 and 6 are maintained in the fingerprint as the sum of 8.

2. Now if we use "nBits=64", there will be no collision when hash those 6 different identifiers.

mol = AllChem.MolFromSmiles('OOOOOOOO')
fp = AllChem.GetMorganFingerprint(mol, 1)
display(UIntSparseIntVect.GetNonzeroElements(fp))

fp = AllChem.GetHashedMorganFingerprint(mol, 1, nBits=64,
  bitInfo=bitInfo)
display(UIntSparseIntVect.GetNonzeroElements(fp))

# output 
{864662311: 2, 864674487: 6, 2379014453: 2, 3705148550: 4, 3706500114: 2}
{6: 4, 18: 2, 39: 2, 53: 2, 55: 6}