Find Clusters in a Set of Molecules

Use molecule fingerprints to group similar molecules together and view them in a feature space plot

Molecule fingerprints are a method for encoding chemical structure in a set of ones and zeros. In this example we use extended connectivity fingerprints to group similar molecules together in clusters.

Install the MoleculeFingerprints paclet and load the package:

In[1]:=

PacletInstall["WolframChemistry/MoleculeFingerprints"];Needs["WolframChemistry`MoleculeFingerprints`"]

Load in a set of SMILES strings included with the paclet for use in examples:

In[2]:=

molecules=Molecule/@ResourceData["Example""Find Clusters in a Set of Molecules","SMILES strings for 300 antibiotic molecules"];MoleculePlot/@RandomSample[molecules,4]

Out[2]=

Molecule fingerprints are a numeric representation of the features in a molecule—encoding the various atom and bond environments in lists of ones and zeros:

In[3]:=

ExtendedConnectivityFingerprint[Echo[First[molecules]]]

Molecule

Formula:

Atoms:

Bonds:



Out[3]=

NumericArray

Type: Integer8

Dimensions: {2048}



In[4]:=

Counts@Normal@%

Out[4]=

02009,139

Use

FeatureSpacePlot

for a qualitative visualization of the distribution of molecules in a set.

FeatureSpacePlot

will call

DimensionReduce

on the molecule's fingerprints to reduce the dimension from 2,048 to 2, plotting each molecule as a point in 2D space. Use tooltips to identify the molecules visually:

In[5]:=

FeatureSpacePlot[molecules,FeatureExtractorExtendedConnectivityFingerprint[],LabelingFunction(Placed[Dynamic[MoleculePlot[#1,ImageSizeSmall]],Tooltip]&)]

Out[5]=

Use

FindClusters

to group the molecules into sets based on fingerprint similarity:

In[6]:=

clusters=FindClusters[molecules,FeatureExtractorExtendedConnectivityFingerprint];Length/@clusters

Out[6]=

{58,10,8,105,29,21,2,2,10,3,3,4,16,2,3,6,3,8,5,2,3,2}

Show the molecules in each cluster in a dynamic slideshow:

In[7]:=

plots=Table[PacletSymbol["WolframChemistry/ChemistryFunctions","MoleculeGridPlot"][Partition[cluster〚;;UpTo[8]〛,2],PlotLabelRow[{"cluster ",n}]],{cluster,clusters}];Manipulate[plots〚n〛,{n,1,Length[clusters],1}]

Out[7]=

Publisher Information

Contributed by: Jason Biggs

Wolfram Language Example Repository

Find Clusters in a Set of Molecules

See Also

Publisher Information

Find Clusters in a Set of Molecules

See Also

Related Symbols

Publisher Information