Cheminformatics Tutorials - Herong's Tutorial Examples - v2.01, by Herong Yang
Impact of 'tgtDensity' on RDKFingerprint()
This section provides a tutorial example on impact of the 'tgtDensity' option on fingerprint generation with rdkit.Chem.rdmolops.RDKFingerprint() function.
The 'tgtDensity' option in the rdkit.Chem.rdmolops.RDKFingerprint() function call allows you to control the folding behavior in the fingerprint generation process. The default value, tgtDensity=0, specifies no folding and related option, minSize, is ignored.
When tgtDensity>0 is specified, the density and the length of the initial fingerprint generated from subgraphs will be examined. If the density is lower than tgtDensity and the length is greater than minSize, the fingerprint will be folded by half. Then the density and the length will be examined again, then fingerprint could be folded again, until the density is >= tgtDensity or the length is <= minSize.
Note that the folding operation is carried out by cutting the fingerprint into 2 bit strings of the same length. Then join them into a single bit string using the bitwise OR operation.
By the way, fingerprint density is defined as the ratio of number of '1' bits over number of all bits.
Having fingerprints with higher density can result shorter fingerprints, and save data storage. But it also reduces the accuracy the final fingerprints.
1. For example, "CCCCCCCC" will have a 64-bit fingerprint without any folding, if "tgtDensity=0.0" is used. The fingerprint density is 7/64 = 0.109375.
from rdkit import Chem
mol = Chem.MolFromSmiles('CCCCCCCC')
fp = Chem.RDKFingerprint(mol, fpSize=64, nBitsPerHash=1,
tgtDensity=0.0, minSize=64)
print(fp.ToBitString())
print(fp.GetNumOnBits()/fp.GetNumBits())
# output
0000000000000100000011000000100001000000000100000000000001000000
0.109375
2. Now if we use "tgtDensity=0.15" and "minSize=16", the fingerprint will be folded once. The length is reduced to 32 and the density is increased to 7/32 = 0.21875.
mol = Chem.MolFromSmiles('CCCCCCCC')
fp = Chem.RDKFingerprint(mol, fpSize=64, nBitsPerHash=1,
tgtDensity=0.15, minSize=16)
print(fp.ToBitString())
print(fp.GetNumOnBits()/fp.GetNumBits())
# output
01000000000101000000110001001000
0.21875
As you can see, there is no loss of information in this folding operation. All 7 "1" bits are maintained.
3. Now let's try molecule "NNNNNNNN". The folding operation will result a loss of 1 bit.
mol = Chem.MolFromSmiles('NNNNNNNN')
fp = Chem.RDKFingerprint(mol, fpSize=64, nBitsPerHash=1,
tgtDensity=0.0, minSize=64)
print(fp.ToBitString())
print(fp.GetNumOnBits()/fp.GetNumBits())
# output
0000001000000000100100100000000000000010000001000000000100000000
0.109375
mol = Chem.MolFromSmiles('NNNNNNNN')
fp = Chem.RDKFingerprint(mol, fpSize=64, nBitsPerHash=1,
tgtDensity=0.15, minSize=16)
print(fp.ToBitString())
print(fp.GetNumOnBits()/fp.GetNumBits())
# output
00000010000001001001001100000000
0.1875
The folding operation is shown below with the lost bit identified.
00000010000000001001001000000000
00000010000001000000000100000000 OR
================================
00000010000001001001001100000000
^ loss
Conclusion: We should keep the default of tgtDensity=0 to stop folding to avoid any loss of information.
Table of Contents
SMILES (Simplified Molecular-Input Line-Entry System)
Open Babel: The Open Source Chemistry Toolbox
Using Open Babel Command: "obabel"
Generating SVG Pictures with Open Babel
Substructure Search with Open Babel
Similarity Search with Open Babel
Fingerprint Index for Fastsearch with Open Babel
Stereochemistry with Open Babel
Command Line Tools Provided by Open Babel
RDKit: Open-Source Cheminformatics Software
rdkit.Chem.rdchem - The Core Module
rdkit.Chem.rdmolfiles - Molecular File Module
rdkit.Chem.rdDepictor - Compute 2D Coordinates
rdkit.Chem.Draw - Handle Molecule Images
Molecule Substructure Search with RDKit
rdkit.Chem.rdmolops - Molecule Operations
►Daylight Fingerprint Generator in RDKit
What Is Daylight Fingerprint Generator in RDKit
RDKFingerprint() Method in RDKit
Impact of 'useBondOrder' on RDKFingerprint()
Impact of 'branchedPaths' on RDKFingerprint()
Impact of 'maxPath' on RDKFingerprint()
Impact of 'fpSize' on RDKFingerprint()
►Impact of 'tgtDensity' on RDKFingerprint()
Impact of 'nBitsPerHash' on RDKFingerprint()
UnfoldedRDKFingerprintCountBased() Method in RDKit
GetRDKitFPGenerator() Method in RDKit
Morgan Fingerprint Generator in RDKit
RDKit Performance on Substructure Search
Introduction to Molecular Fingerprints
OCSR (Optical Chemical Structure Recognition)
AlphaFold - Protein Structure Prediction