Protein targets are listed on PubChem BioAssay record's as they are provided by submitters. There are several work-arounds that need to be performed to ensure all data is extracted.
Using Eutils, we link from UniProt Accessions to Genes (or start with EGIDs) and identical protein products for that Uniprot (even the ones not from UniProt). This way, regardless of whether the submitter used UniProt IDs or GenPept IDs or RefSeq IDs or GeneIDs, we are able to find BioAssays with a defined target the same as indicated by the UniProt accession. We then link to the active or inactive compounds from the BioAssays. We retrive the SMILES for the unique CIDs and map those back to the links between the compounds and Uniprot, and finally calculate the fingerprints for the SMILES.
This method uses the following resources:- 'Eutils: Entrez Programming Utilities' (http://www.ncbi.nlm.nih.gov/books/NBK25501/) 'PubChem PUG'(https://pubchem.ncbi.nlm.nih.gov/pug/pughelp.html)
This script is uses threading in order to achieve an acceptable completion time
Dependencies : 'lxml', 'rdkit', 'Chemaxon JChem', 'OrderedDict'
Options:
-h, --help show this help message and exit
-i FILE, --input=FILE
Input file (Uniprots [default] or EGIDs [must supply
the --geneids option])
--geneids Toggle Entrez Gene ID input format
--actives Toggle extract actives
--inactives Toggle extract inactives