Search

Adverse Drug Reactions (ADRs) are characterized within randomized clinical trials and postmarketing pharmacovigilance, but their molecular mechanism remains unknown in most cases. Aside from clinical trials, many elements of knowledge about drug ingredients are available in open-access knowledge graphs. In addition, drug classifications that label drugs as either causative or not for several ADRs, have been established. We propose to mine knowledge graphs for identifying biomolecular features that may enable reproducing automatically expert classifications that distinguish drug 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, we mine a knowledge graph for features; we train classifiers at distinguishing, drugs associated or not with ADRs; 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 we manually evaluate how they may be explanatory. Extracted features reproduce with a good fidelity classifications of drugs causative or not for DILI and SCAR. 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. Knowledge graphs provide 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.

International audience; Overcoming the problem of resistance to antiretroviral drugs (ARVs) in HIV-infected patients is a major issue in AIDS research today. Advances in genome sequencing have facilitated the identification of a growing number of individual genotypes. Hence, it is now possible to understand HIV drug resistance at the molecular level by considering the three-dimensional (3D) structural interactions between ARVs and the mutated viral proteins of patients. Therefore, identification of the critical interactions lost further to one or several HIV mutations, and consequently the modifications of other molecular factors, could be indicators to propose appropriate ARVs escaping the resistance. This paper introduces the HIV-PDI (Protein-Drug Interactions) resource designed to be a decision making tool to propose alternative ARVs against a particular mutated viral protein, and thus to provide a personalized antiretroviral treatment. The HIV-PDI was conceived to serve as an integrated resource for studying HIV drug resistance at the structural level of the protein-drug interaction, with a special emphasis on the active site of the HIV drug target. As a first step, we focus on the well documented protease and related drugs. The HIV-PDI includes clinical information on patients, resistance to given ARVs treatments, HIV proteins structures and mutations, HIV protein/ARV drugs and their 3D interactions. The HIV-PDI may be queried using multiple combinations of fields including protein, drug and treatment conditions and coupled to visualization/analysis tools of 3D Protein-Drug interactions. The HIV-PDI resource can be used in order to help understand the appearance of resistance and to promote further novel drug and treatment developments based on analyses of 3D pattern of protein-drug interactions. A web-based version of HIV-PDI is available at http://hiv-pdi.loria.fr.

Bioinformatic data sources available on the web are multiple and heterogenous. The lack of documentation and the difficulty of interaction with these data sources require users competence in both informatics and biological fields for an optimal use of sources contents that remain rather under exploited. In this paper we present an approach based on formal concept analysis to classify and search relevant bioinformatic data sources for a given query. It consists in building the concept lattice from the binary relation between bioinformatic data sources and their associated metadata. The concept built from a given query is then merged into the concept lattice. The result is given by the extraction of the set of sources belonging to the extents of the query concept subsumers in the resulting concept lattice. The sources ranking is given by the concept specificity order in the concept lattice. An improvement of the approach consists in automatic query refinement thanks to domain ontologies. Two forms of refinement are possible by generalisation and by specialisation. ----- Les sources de donn��es biologiques disponibles sur le web sont multiples et h��t��rog��nes. L'utilisation optimale de ces ressources n��cessite aujourd'hui de la part des utilisateurs des comp��tences �� la fois en informatique et en biologie, du fait du manque de documentation et des difficult��s d'interaction avec les sources de donn��es. De fait, les contenus de ces ressources restent souvent sous-exploit��s. Nous pr��sentons ici une approche bas��e sur l'analyse de concepts formels, pour organiser et rechercher des sources de donn��es biologiques pertinentes pour une requ��te donn��e. Le travail consiste �� construire un treillis de concepts �� partir des m��ta-donn��es associ��es aux sources. Le concept construit �� partir d'une requ��te donn��e est alors int��gr�� au treillis. La r��ponse �� la requ��te est ensuite fournie par l'extraction des sources de donn��es appartenant aux extensions des concepts subsumant le concept requ��te dans le treillis. Les sources ainsi retourn��es peuvent ��tre tri��es selon l'ordre de sp��cificit�� des concepts dans le treillis. Une proc��dure de raffinement de requ��te, bas��e sur des ontologies du domaine, permet d'am��liorer le rappel par g��n��ralisation ou par sp��cialisation