Skip to content

Steinbeck-Lab/COCONUT-Legacy

 
 

Repository files navigation

Compiler for the COlleCtion of Open NatUral producTs (COCONUT)

This compiles is designed to read molecules from various file types (SMILES, SDF, MOL, csv, tsv), check them for errors and connectivity, compute a large number of molecular parameters and properties, and store everything in a Mongo database. Unless you want to modify the code, we recomment using the compiled JAR that can be downloaded here: https://zenodo.org/record/3695455

System pre-requisites:

  • MongoDB installed and accessible by the default (27017) port on localhost
  • Java minimum 8 version installed

Load COCONUT

You can download the latest version COCONUT from ZENODO (https://zenodo.org/record/3688734). In case you want to explore the whole database in MongoDB, you can load the downloaded dataset dump as following:

unzip COCONUT2020-07.zip
cd COCONUT2020-07/COCONUT2020-07/
mongorestore --db=COCONUT --noIndexRestore .

Note that for system compatibilities reason, it is better to restore the database without the indexes (hence the "noIndexRestore" option). However, seen the size of the dataset, we suggest to add the indexes as following:

mongo
use COCONUT2020-07


 db.sourceNaturalProduct.createIndex( {source:1})

 db.sourceNaturalProduct.createIndex( {simpleInchi:"hashed"})

 db.sourceNaturalProduct.createIndex( {simpleInchiKey:1})
 db.sourceNaturalProduct.createIndex( {originalInchiKey:1})
 db.sourceNaturalProduct.createIndex( {originalSmiles:"hashed"})
 db.sourceNaturalProduct.createIndex( {absoluteSmiles:"hashed"})
 db.sourceNaturalProduct.createIndex( {idInSource:1})


db.uniqueNaturalProduct.createIndex( {inchi:"hashed"})
db.uniqueNaturalProduct.createIndex( {inchikey:1})
db.uniqueNaturalProduct.createIndex( {clean_smiles: "hashed"})
db.uniqueNaturalProduct.createIndex( {molecular_formula:1})
db.uniqueNaturalProduct.createIndex( {coconut_id:1})
db.uniqueNaturalProduct.createIndex( {fragmentsWithSugar:"hashed"})
db.uniqueNaturalProduct.createIndex( {fragments:"hashed"})
db.uniqueNaturalProduct.createIndex( {annotationLevel:1})
db.uniqueNaturalProduct.createIndex( {synonyms:"text", name:"text"})
db.uniqueNaturalProduct.createIndex( {npl_score:1})

db.uniqueNaturalProduct.createIndex( { pubchemBits : "hashed" } )

db.uniqueNaturalProduct.createIndex( {unique_smiles: "hashed"})

db.uniqueNaturalProduct.createIndex( { "pfCounts.bits" :1} )
db.uniqueNaturalProduct.createIndex( { "pfCounts.count" : 1 })


db.fragment.createIndex({signature:1})
db.fragment.createIndex({signature:1, withsugar:-1})



Required folder structure

COCONUT
├── coconut-0.0.1-SNAPSHOT.jar # the compiled jar. It can be downloaded from ZENODO: https://zenodo.org/record/3695455
├── coconut_ids_june2020.csv
├── data # here go the files with NP molecular structures
├── UpdateCOCONUT
│   ├── COCONUTupdater.py
│   ├── VerifyNames.py
├── fragments
│   ├── fragment_without_sugar.txt
│   ├── fragment_with_sugar.txt
├── sm # file(s) with structures of synthetic molecules 

Execution options

Run COCONUT compilation from scratch
java -Xmx16288m -jar coconut-0.0.1-SNAPSHOT.jar data sm/sm.tsv fragments/fragment_without_sugar.txt fragments/fragment_with_sugar.txt importCOCONUTids coconut_ids_june2020.csv > logs.txt &
Re-run COCONUT to recompute missing molecular features
java -jar coconut-0.0.1-SNAPSHOT.jar recomputeMissing &
Run only similarity calculation

The provided script also computes similatiry between molecules in COCONUT. This step requires a large amount of memory (this will be optimized in the future) and might need a separate calculation.

java -Xmx12288m -jar coconut-0.0.1-SNAPSHOT.jar runOnlySimilarity &
Run only addition of synthetic molecules

Synthetic molecules (SM) are required for a large number of comparisons with NPs (for example a re-calculation from scratch of NP-likeness score). You need to provide your own dataset of synthetic molecules (SM), we suggest using the ZINC15 dataset. Note that not adding any SM will not affect COCONUT, unless there is a need of re-calculating the NP-likeness score from scratch.

java -jar coconut-0.0.1-SNAPSHOT.jar onlyAddSM ~/Projects/NP/COCONUT/sm/sm.tsv & 
Run only CNPid calculation

In case you need to produce your own CNPid (COCONUT NP identifiers) - not recommened.

java -jar coconut-0.0.1-SNAPSHOT.jar addCNPid  & 
run only import of CNPid from file

java -jar coconut-0.0.1-SNAPSHOT.jar onlyImportCoconutIds coconut_ids_june2020.csv &

Molecule names curation

Import names from ChEBI, PubChem and CMAUP
python3 UpdateCOCONUT/COCONUTupdater.py
Import IUPAC names and clean
python3 UpdateCOCONUT/VerifyNames.py
Run annotation level recalculation
java -Xmx16288m -jar coconut-0.0.1-SNAPSHOT.jar evaluateAnnotation &
Names to low case
java -Xmx16288m -jar coconut-0.0.1-SNAPSHOT.jar namesToLowerCase &

Releases

No releases published

Packages

No packages published

Languages

  • Java 68.6%
  • Jupyter Notebook 11.5%
  • HTML 10.3%
  • Python 9.6%