uc_FIDO: unambiguous characterization of food interactions with drugs ontology Constantine W. Spyrou Matthew C. Lange, PhD. Dept. of Food Science Dept. of Food Science UC Davis UC Davis Davis, CA Davis, CA cwspyrou@ucdavis.edu mclange@ucdavis.edu Abstract— uc_FIDO is an ontology that unambiguously discovering and understanding all pharmacological, nutritional, characterizes food interactions with drugs in the human body. and drug efficacy effects of these interactions. This ontology is part of a group of food ontologies describing food and the human experience at the International Center for uc_FIDO is designed as part of IC-FOODS, a larger Food Ontology Operability, Data and Semantics (IC-FOODS) at network of unambiguously characterized food ontologies UC Davis. The first of its kind, uc_FIDO characterizes relations meant to connect bridges between food systems, food and between food, medicine, and human health. uc_FIDO brings health [4]. together several existing ontologies related to anatomy, metabolic pathways, biological processes, drug ingredients and food II. DESIGN AND METHODS structures. Through these ontologies, uc_FIDO annotates relationships between food and drug bioactives, human A. Scope and Knowledge Elicitation physiological conditions, and biological reaction pathways. Relationships that link together fully characterize various food uc_FIDO encompasses multiple ontologies and exclusively interactions with drugs and their effects. contains nearly forty thousand axioms and over two thousand classes. Ontologies are integrated via Protégé 5.0.0.24 and The current dearth of ontologies for characterizing foods updated versions of uc_FIDO are uploaded to GitHub. Within limits advancement of informatics solutions for improving health. uc_FIDO, food and drug interaction modelers can utilize As ontologies of foods are developed, it becomes necessary to various base ontologies. For drug active ingredients and describe ingredients, bioactive molecules, potential toxins, and chemical interactions, the Drug Ontology [5], [6] CheBI and other molecules in food interacting with drugs and the human ingredient sub-ontologies provide a foundation to describe body. drugs’ effects and active ingredients. uc_Eating [7] defines eating behaviors for various situations, which is crucial to Keywords—bioactive; interaction; metabolic process; active establishing a basis of foods in the human body because drug ingredient; food component; food matrix psychological behavior affects food selection [8]. I. INTRODUCTION (MOTIVATION AND INTENDED USAGE) Multiple linkages formed within base ontologies create different food-drug interactions. Each food-drug interaction in Approximately seven percent of hospitalizations are the the ontology maps sub-classes relating to food bioactive result of adverse drug reactions [1], many of which occur ingredients, drug bioactives, and human body pathways that between drugs and food. Past research clarifies these make up food-drug interactions. Interactions, bioactives, and potentially deadly reactions between our food and medicine. pathways are the primary classes that link together to build and Examples include interactions between grapefruit and drugs characterize interactions. Subclasses that have properties which alter pathways involving cytochrome P450 (CytP450). linking to each other create defined interactions and serve as Drugs such as Lipitor (atorvastatin, a cholesterol-lowering primary classes within uc_FIDO. drug) were found in dangerously high concentrations in the bloodstream when consumed with grapefruit juice [2]. Drugs Active ingredients primarily compose the class also limit nutrient absorption from many foods. hierarchy drawn from drug and food ontologies. However, Corticosteroids, for example, are linked to increased calcium because uc_FIDO is intended as a knowledge source for excretion from the body [3]. These examples merely highlight consumer-friendly tools, features like brands become necessary the broad array of effects and sources involved in food-drug to define as sub-classes. Because of the multiple properties of interactions. Much research has been undertaken, yet no different foods based on food matrices, concentrations of knowledge repository captures the scope of food-drug related nutrients in different ingredients (ie. Skim milk vs. whole interactions, adverse, favorable, or otherwise. uc_FIDO milk), each food becomes a specific class with multiple provides a platform upon which development of ecosystems of subclasses relating to that food. For example, the class consumer tools surrounding foods, drugs, and education can be “sourdough” would have subclasses relating to ingredients built to unify information resources. This will be crucial in (“all-purpose_flour”), recipes, bioactive molecules, and nutrition content. Bioactive molecules and nutrients react with drug bioactives in the human body, and thus are crucial to classes of interactions in base ontologies. Figure 2 below mapping out defined interactions. shows some subclasses considered because of the possible The location and stimuli of human body reactions require locations of food-drug interactions. full comprehension. Understanding these is crucial to building uc_FIDO. Specifically, most reactions occur in the blood stream, the blood-brain barrier, and in organs. Defining these reactions, nutrient cycles, and biological pathways that are affected by these potential interactions is crucial. Biological pathways are much more complex based on anatomy and location. uc_FIDO draws from Uberon as a class resource for anatomical structures and locations, while Reactome integrates base ontologies of biological pathways [9]. From here, human body reaction pathways like “sensory_process” or “mechanical_food_breakdown” and anatomical locations including “lung”, “alveoli”, and others, become key classes that link to where drug and food bioactive compounds are Figure 2. Examples of subclasses developed as part of the interaction class hierarchy for uc_FoodDrugInteractionOntology. metabolized by physiological location and anatomical structures where reactions occur. Through properties like physiological location and common reactants or products, Linkages of these subclasses occur through sharing interactions are defined by linking together different bioactives properties including anatomical location and reactant or and conditions. For example, the class product molecules. The sharing of properties between “bloodstream_interaction” links to the classes “Lipitor” and interactions, anatomical structures, and bioactive pathway “Grapefruit” through sharing the property “has_location sites allows for accurate mapping of where and how these bloodstream”. interactions occur, along with their effects. B. Ontology Mapping III. RESULTS AND ANALYSIS Mapping of uc_FIDO was undertaken similarly to The completion of uc_FIDO requires base ontologies Joslin’s One-Carbon Metabolism map [10]. CMAPs currently under construction. Specifically, ontologies initially mapped food drug interactions split into three surrounding food need to be more developed. While several ontologies regarding food are progressing [11], no current food components – food, drugs, and body reaction pathways. ontology takes into account all different types of bioactive compounds found in foods. Ingredients and food additives have been taken into account in some cases, but pesticides, natural food toxins, and other bioactives also need to be considered as they may have effects on drug efficacy or pharmacological effects. Additional factors (ie. other supplements or foods consumed) also factor into drug concentration bloodstreams, contributing to a network of multiple simultaneous interactions that requires further research. uc_FIDO is a great start for mapping relationships between drugs and other substances intravenous or extraneous to the human body, such as food. However, more ontologies need development to describe the relationship between nutrients and genes [10] or ontologies that understand how food and nutrients affect biological pathways and human reactions, along with other food related ontologies. When these ontologies are completed, actions of food and drugs within human bodies can be realized in a clean, easy-to- Figure 1. Concept Map generalizing but detailing components of use ontology. human body reaction pathways in uc_FIDO. While uc_FIDO continues to be developed under IC- Figure 1 above shows CMAPs effectively creating FOODS, ontologies described in the above paragraphs relating generalized previews of uc_FIDO. Each class of human to all chemicals present in food and nutrient-gene interactions body reactions above has subclasses like disease states, (among others) can be developed. When all IC-FOODS concentrations, and pH levels that all factor into whether ontologies are completed, they can power multiple consumer reactions within food drug interactions will take place. and medical Internet applications that inform and promote quicker responses to food emergencies while increasing access to food education information. When translating the concept map into linkages and classes in Protégé, many extra subclasses are considered. Because of several potential sites of food-drug interactions [11], these must be taken into account when mapping out REFERENCES [6] Unknown Author, “DrON Ingredients Ontology.” Unpublished. [7] Kimiya Taji and Matthew C. Lange, “uc_Eating: Ontology for [1] Jason Lazarou, Bruce H. Pomeranz, and Paul N. Corey, “Incidence of unambigious characterization of eating and food habits.” Adverse Drug Events in Hospitalized Patients: A Meta-Analysis of Unpublished., 01-Jul-2016. Prospective Studies,” J. Am. Med. Assoc., vol. 279, no. 15, pp. 1200– [8] F Bellisle, “Why should we study human food intake behaviour?,” 1205, May 1998. Nutr. Metab. Cardiovasc. Dis., vol. 13, no. 4, pp. 189–193, Aug. [2] Irina Piatkov, Trudi Jones, and Mark McLean, “Drug Interactions, 2003. Pharmacogenomics and Cardiovascular Complication,” Intech, 2013. [9] “Reactome BioPathway Software.” Unpublished. [3] RM Carney, KE Freedland, EH Rubin, MW Rich, BC Steinmeyer, [10] A. C. Joslin, R. Green, J. B. German, and M. C. Lange, “Concept and WS Harris, “Omega-3 augmentation of sertraline in treatment of mapping One-Carbon Metabolism to model future ontologies for depression in patients with coronary heart disease: a randomized nutrient–gene–phenotype interactions,” Genes Nutr., vol. 9, no. 5, controlled trial.,” J. Am. Med. Assoc., vol. 302, no. 15, pp. 1651–1657, Sep. 2014. Oct. 2009. [11] Maged N. Kamel Boulos, Abdulslam Yassine, Shervin [4] Matthew C Lange, Danielle G Lemay, and J Bruce German, “A Multi- Shirmohammadi, Chakkrit Snae Namahoot, and Michael Bruckner, ontology Framework to Guide Agriculture and Food towards Diet and “Towards an ‘Internet of Food’: Food Ontologies for the Internet of Health,” J. Sci. Food Agric., vol. 87, pp. 1427–1434, Apr. 2007. Things,” Future Internet, vol. 7, pp. 372–392, Oct. 2015. [5] Unknown Author, “DrON Chebi.” Unpublished., 2015.