Emmanuel Bresso1,2, Pierre Monnin1,3, Cédric Bousquet4,5, François-Elie Calvier4, Ndeye-Coumba Ndiaye6, Nadine Petitpain7, Malika Smaïl-Tabbone1, Adrien Coulet8,9,10. 1. Université de Lorraine, CNRS, Inria, LORIA, Nancy, France. 2. Centre d'Investigations Cliniques Plurithématique 1433, Inserm 1116, CHRU de Nancy, Université de Lorraine, Nancy, France. 3. Orange, Belfort, France. 4. Service de santé publique et information médicale, CHU de Saint Etienne, Saint Etienne, France. 5. Sorbonne Université, Inserm, Université Paris 13, LIMICS, Paris, France. 6. Université de Lorraine, Inserm U1256 - NGERE, Nancy, France. 7. Centre Régional de Pharmacovigilance, CHRU of Nancy, Nancy, France. 8. Université de Lorraine, CNRS, Inria, LORIA, Nancy, France. adrien.coulet@inria.fr. 9. Inria Paris, Paris, France. adrien.coulet@inria.fr. 10. Centre de Recherche des Cordeliers, INSERM, Sorbonne Université, Université de Paris, Paris, France. adrien.coulet@inria.fr.
Abstract
BACKGROUND: Adverse drug reactions (ADRs) are statistically characterized within randomized clinical trials and postmarketing pharmacovigilance, but their molecular mechanism remains unknown in most cases. This is true even for hepatic or skin toxicities, which are classically monitored during drug design. Aside from clinical trials, many elements of knowledge about drug ingredients are available in open-access knowledge graphs, such as their properties, interactions, or involvements in pathways. In addition, drug classifications that label drugs as either causative or not for several ADRs, have been established. METHODS: We propose in this paper to mine knowledge graphs for identifying biomolecular features that may enable automatically reproducing expert classifications that distinguish drugs causative or not for a given type of ADR. In an Explainable AI perspective, we explore simple classification techniques such as Decision Trees and Classification Rules because they provide human-readable models, which explain the classification itself, but may also provide elements of explanation for molecular mechanisms behind ADRs. In summary, (1) we mine a knowledge graph for features; (2) we train classifiers at distinguishing, on the basis of extracted features, drugs associated or not with two commonly monitored ADRs: drug-induced liver injuries (DILI) and severe cutaneous adverse reactions (SCAR); (3) we isolate features that are both efficient in reproducing expert classifications and interpretable by experts (i.e., Gene Ontology terms, drug targets, or pathway names); and (4) we manually evaluate in a mini-study how they may be explanatory. RESULTS: Extracted features reproduce with a good fidelity classifications of drugs causative or not for DILI and SCAR (Accuracy = 0.74 and 0.81, respectively). Experts fully agreed that 73% and 38% of the most discriminative features are possibly explanatory for DILI and SCAR, respectively; and partially agreed (2/3) for 90% and 77% of them. CONCLUSION: Knowledge graphs provide sufficiently diverse features to enable simple and explainable models to distinguish between drugs that are causative or not for ADRs. In addition to explaining classifications, most discriminative features appear to be good candidates for investigating ADR mechanisms further.
BACKGROUND: Adverse drug reactions (ADRs) are statistically characterized within randomized clinical trials and postmarketing pharmacovigilance, but their molecular mechanism remains unknown in most cases. This is true even for hepatic or skin toxicities, which are classically monitored during drug design. Aside from clinical trials, many elements of knowledge about drug ingredients are available in open-access knowledge graphs, such as their properties, interactions, or involvements in pathways. In addition, drug classifications that label drugs as either causative or not for several ADRs, have been established. METHODS: We propose in this paper to mine knowledge graphs for identifying biomolecular features that may enable automatically reproducing expert classifications that distinguish drugs causative or not for a given type of ADR. In an Explainable AI perspective, we explore simple classification techniques such as Decision Trees and Classification Rules because they provide human-readable models, which explain the classification itself, but may also provide elements of explanation for molecular mechanisms behind ADRs. In summary, (1) we mine a knowledge graph for features; (2) we train classifiers at distinguishing, on the basis of extracted features, drugs associated or not with two commonly monitored ADRs: drug-induced liver injuries (DILI) and severe cutaneous adverse reactions (SCAR); (3) we isolate features that are both efficient in reproducing expert classifications and interpretable by experts (i.e., Gene Ontology terms, drug targets, or pathway names); and (4) we manually evaluate in a mini-study how they may be explanatory. RESULTS: Extracted features reproduce with a good fidelity classifications of drugs causative or not for DILI and SCAR (Accuracy = 0.74 and 0.81, respectively). Experts fully agreed that 73% and 38% of the most discriminative features are possibly explanatory for DILI and SCAR, respectively; and partially agreed (2/3) for 90% and 77% of them. CONCLUSION: Knowledge graphs provide sufficiently diverse features to enable simple and explainable models to distinguish between drugs that are causative or not for ADRs. In addition to explaining classifications, most discriminative features appear to be good candidates for investigating ADR mechanisms further.
Entities:
Keywords:
Adverse drug reaction; Data mining; Explainable AI; Explanation; Knowledge graph; Machine learning; Mechanism of action; Molecular mechanism
Authors: Gianluca Trifirò; Antoine Pariente; Preciosa M Coloma; Jan A Kors; Giovanni Polimeni; Ghada Miremont-Salamé; Maria Antonietta Catania; Francesco Salvo; Anaelle David; Nicholas Moore; Achille Patrizio Caputi; Miriam Sturkenboom; Mariam Molokhia; Julia Hippisley-Cox; Carlos Diaz Acedo; Johan van der Lei; Annie Fourrier-Reglat Journal: Pharmacoepidemiol Drug Saf Date: 2009-12 Impact factor: 2.890