The French Agency for the Safety of Health Products (ANSM) publishes guidelines on Potential Drug Drug Interaction (PDDI) at this address: https://ansm.sante.fr/documents/reference/thesaurus-des-interactions-medicamenteuses-1
A program extracts the content of PDF documents and structures them in JSON format: https://github.com/scossin/ExtractThesaurusANSM The PDF files are available in this repository.
This package provides an API to query the guidelines, it is used to detect PDDI according to French guidelines.
This is a screenshot of a PDDI in the ANSM document:
Colistine is a substance. It can interact with Curares, a drug class. In this example, Colistine interacts on his own, i.e. this interaction is specific to this substance.
Colistine belongs to the médicaments néphrotoxiques drug class. So when looking for PDDIs with Colistine, one must also have a look to PDDIs related to the médicaments néphrotoxiques drug class.
The interaction mechanism is:
Potentialisation des curares lorque l'antibiotique est administré par nvoie parentérale et/ou péritonéale avant, pendant ou après l'agent ncurarisant. (Potentiation of curares when the antibiotic is administered parenterally parenteral and/or peritoneal route before, during or after the curarizing agent.)
The severity level is Précaution d'emploi. There are 4 severity levels in the document:
- Contraindicated (contre-indiquée)
- Discommended (association déconseillée)
- Needing cautious use (précaution d'emploi)
- To take into account (à prendre en compte)
Most PDDIs have one severity level. Some PDDIs have multiple severity levels which can depend on quantity of drugs administered or patient medical history.
Often the document provides information about the severity level. In this example:
Surveiller le degré de curarisation en fin d'anesthésie (Monitor the degree of curarization at the end of anesthesia)
The JSON representation of this PDDI is the following:
{
"main_drug": "COLISTINE",
"between_main_and_plus_drug": "Voir aussi : médicaments néphrotoxiques",
"plus_drug": "CURARES",
"severity_levels": [
{
"level": "Précaution d'emploi",
"info": "Surveiller le degré de curarisation en fin d'anesthésie."
}
],
"interaction_mechanism": "Potentialisation des curares lorque l'antibiotique est administré par \nvoie parentérale et/ou péritonéale avant, pendant ou après l'agent \ncurarisant.",
"description": "Précaution d'emploi\nSurveiller le degré de curarisation en fin d'anesthésie.\n\nPotentialisation des curares lorque l'antibiotique est administré par \nvoie parentérale et/ou péritonéale avant, pendant ou après l'agent \ncurarisant."
}The severity_levels and the interaction_mechanism are extracted automatically from the description field. This field is not useful but to verify information was correctly structured.
Also, a "Substance file" stores information about substances and their drug classes. The JSON colistine entry is the following:
{
"substance": "colistine",
"drug_classes": [
"médicaments néphrotoxiques",
"autres médicaments néphrotoxiques"
]
}- Check PDDI between colistine and curares
The first step is to select a thesaurus_version. The ANSM interaction working group publishes guidelines once or twice a year. The PDDIs detected depend on the selected version.
from pddiansm.detector.PDDIthesaurusDetector import PDDIthesaurusDetector from pddiansm.thesaurus.ThesauriJson import ThesauriJson ThesauriJson().print_available_thesaurus_version() thesaurus_version = "2019_09" thesaurus = ThesauriJson().get_thesaurus(thesaurus_version) pddi_detector = PDDIthesaurusDetector(thesaurus) substance1 = "colistine" substance2 = "curares" pddis = pddi_detector.detect_pddi(substance1, substance2) for pddi in pddis: print(pddi) # colistine (from 'colistine') can interact with curares (from 'curares') in thesaurus version 2019_09
- Check PDDI between colistine and mivacurium
Since mivacurium belongs to the curares drug class, the former PDDI is also detected between these two substances:
substance1 = "colistine" substance2 = "mivacurium" pddis = pddi_detector.detect_pddi(substance1, substance2) for pddi in pddis: print(pddi) # colistine (from 'colistine') can interact with mivacurium (from 'curares') in thesaurus version 2019_09
- Check PDDI between two simple drugs
A simple drug is an object, with an id, containing one to many substances. Marketed drugs can contain one to several substances, the PDDIsimpleDrugsDetector class is used to detect PDDIs between substances of different drugs.
from pddiansm.detector.PDDIsimpleDrugsDetector import PDDIsimpleDrugsDetector from pddiansm.detected.PDDIsimpleDrugsDetected import PDDIsimpleDrugsDetected from pddiansm.pydantic.interfaces_input import SimpleDrug from pddiansm.thesaurus.IThesaurus import IThesaurus from pddiansm.thesaurus.ThesauriJson import ThesauriJson simple_drug1 = SimpleDrug(id=1, substances=["opium", "tiemonium", "colchicine"]) simple_drug2 = SimpleDrug(id=2, substances=["azithromycine"]) simple_drugs = [simple_drug1, simple_drug2] thesaurus: IThesaurus = ThesauriJson().get_thesaurus("2019_09") pddi_detector = PDDIsimpleDrugsDetector(thesaurus) pddis_detected = pddi_detector.detect_pddi_multiple_drugs(simple_drugs) for pddi_detected in pddis_detected: print(pddi_detected) # colchicine (from 'colchicine') can interact with azithromycine (from 'macrolides (sauf spiramycine)') in thesaurus version 2019_09. # colchicine comes from drug number '1' and azithromycine comes from drug number '2'
- Check PDDI between two identifiers
You can replace substance names by identifiers and search PDDIs with these identifiers. You need to change the mapper object that maps an identifier to one or many substance(s) and drug classe(s). This package contains mappings to Wikidata and RxNorm. For example, to detect PDDIs between two RxNorm ingredients:
from pddiansm.detector.PDDIthesaurusDetector import PDDIthesaurusDetector from pddiansm.thesaurus.ThesauriJson import ThesauriJson from pddiansm.detected.PDDIdetected import PDDIdetected from pddiansm.mapper.AvailableIdentifierMappers import AvailableSubstanceMapping thesaurus = ThesauriJson().get_thesaurus("2019_09") pddi_detector = PDDIthesaurusDetector(thesaurus) rxnorm_mapper = AvailableSubstanceMapping.RxNorm.value pddi_detector.set_mapper(rxnorm_mapper) id_rxnorm_colchicine = "2683" id_rxnorm_azithromycine = "18631" pddis_detected = pddi_detector.detect_pddi(id_rxnorm_colchicine, id_rxnorm_azithromycine) for pddi in pddis_detected: print(pddi) # 2683 (from 'colchicine') can interact with 18631 (from 'macrolides (sauf spiramycine)') in thesaurus version 2019_09
A thesaurus object contains a list of PDDIs and substances (with their drug classes)
from pddiansm.detector.PDDIthesaurusDetector import PDDIthesaurusDetector from pddiansm.thesaurus.ThesauriJson import ThesauriJson ThesauriJson().print_available_thesaurus_version() thesaurus_version = "2019_09" thesaurus = ThesauriJson().get_thesaurus(thesaurus_version) len(thesaurus.get_pddis()) # this version contains 3,055 pddis len(thesaurus.get_substances_thesaurus()) # and 1,052 substances
To get all the drug classes of colistine:
filter_colistine = filter(lambda x: x.substance == "colistine", thesaurus.get_substances_thesaurus()) list(filter_colistine)[0].drug_classes # ['medicaments nephrotoxiques', 'autres medicaments nephrotoxiques']
.. toctree:: :maxdepth: 2 :caption: Contents: