rdkit/Contrib/Fastcluster

FastCluster

This is simple workflow for clustering molecules

• Author: iwatobipen
• Date: 201712010
• Current version uses Morgan FP; rad2 is used for clustering

Requirements

• python3.x
• bayon https://github.com/fujimizu/bayon
• rdkit

Description

• Users need to install bayon at first, and can find tutorial at following URL. https://github.com/fujimizu/bayon/wiki/Tutorial_English
• Also it is needed to install RDKit for parsing SMILES.
• That's all!

Basic usage

• input file format is tab delimited text format, "ID" \t "SMILES" \n .....
• $ python fastcluster.py {input; inputfile} {N; number of clusters} { --output; filename of output} {--centroid; filename of centroid information}
• Example usage
• $ ptyhon fastcluster.py cdk2.smi 5 # clustering 47 compounds to 5 clusters.

Fastcluster iwatobipen$ python fastcluster.py cdk2.smi 5

real	0m0.015s
user	0m0.006s
sys	0m0.002s

Output format

• clusterd.tsv is default output format of bayon. List of clusters with similarity points.

cluster_1 \t molid1 \t point \t molid2 \t point ... \n
cluster_2 \t molid4 \t point \t molid5 \t point ... \n
....

• cluser_parse.tsv is rectangle format of cluster.tsv

molid1 \t point \t clusterID1 \n
molid2 \t point \t clusterID2 \n
molid3 \t point \t clusterID3 \n
molid4 \t point \t clusterID4 \n
....

Memo

• It will need more cpu time compared with directly using bayon. Because this script converts smiles to fingerprint dataset at first then performs clustering.