chem-usecase

Chemical Similarity Usecase

Background

Chemical Similarity usecase uses Tanimoto algorithm to calculate % similarity between 1 molecule with others. We need to scan the whole data set to filter closest molecules, which inputs are molecule and threshold. For example, given a molecule with SMILE as "IC=C1/CCC(C(=O)O1)c2cccc3ccccc23" and threshold = 90%, returns set of molecules that have similarity >= 90%

Requirements

• RDKit Python: http://www.rdkit.org/docs/Install.html
• ChemBL Dataset: ftp://ftp.ebi.ac.uk/pub/databases/chembl/ChEMBLdb/releases/
• Pilosa: https://github.com/pilosa/pilosa

Import dataset to pilosa

Export data from sdf or mysql to csv file, then import csv to pilosa. Since chembl_id in SD file always come with CHEMBL, e.g CHEMBL6329, which pilosa hasn’t support string id yet, we will remove CHEMBL and import chembl_id with integer after that. If you export data from mysql, it uses molregno in compound_structures as unique integer ID for molecule, so you just need to import molregno as chembl_id

• Export to csv from sdf file: python import_from_sdf -p ~/Downloads/chembl_22.sdf -file id_fingerprint.csv

• Export to csv from mysql: python import_from_mysql -d chembl_22 -u root -file id_fingerprint.csv

• Import from csv file to pilosa (mol and inverse-mol are default database, mole.n is default frame)

  • Open Pilosa server:

     pilosa server
    

  • Create mole database

     curl localhost:10101/index/mole -X POST -d '{"options": {"columnLabel": "position_id"}}'

  • Create frame fingerprint

      curl localhost:10101/index/mole/frame/fingerprint -X POST -d '{"options": {"rowLabel": "chembl_id", "inverseEnabled": true, "cacheSize": 2000000}}'

  • Run import script

     pilosa import -d mole -f fingerprint id_fingerprint.csv

Queries

• Retrieve molecule_ids that have similarity >= 90%. (Default db="mole", frame="fingerprint", hosts=127.0.0.1:10101)

  python similar.py -s "I\C=C/1\CCC(C(=O)O1)c2cccc3ccccc23" -t 90
  

• Retrive molecule_id from a smile

  python get_mol_fr_smile.py -s "I\C=C/1\CCC(C(=O)O1)c2cccc3ccccc23"
  

• Benchmark running for thresholds = [50, 70, 75, 80, 85, 90]

  python benchmarks -id 24
  

License

chem-usecase by Linh Vo

To the extent possible under law, the person who associated CC0 with pilosa chem-usecase has waived all copyright and related or neighboring rights to chem-usecase.

You should have received a copy of the CC0 legalcode along with this work. If not, see http://creativecommons.org/publicdomain/zero/1.0/.