Section: New Results

Knowledge Discovery in Healthcare and Life Sciences

Participants : Miguel Couceiro, Adrien Coulet, Amedeo Napoli, Chedy Raïssi, Mohsen Sayed, Malika Smaïl-Tabbone, Yannick Toussaint.

Life Sciences constitute a challenging domain for KDDK. Biological data are complex from many points of views, e.g. voluminous, high-dimensional and deeply inter-connected. Analyzing such data is a crucial issue in healthcare, environment and agronomy. Besides, many bio-ontologies are available and can be used to enhance the knowledge discovery process. Accordingly, the research work of the Orpailleur team in KDDK applied to Life Sciences is in concern with the use of bio-ontologies to improve KDDK, and as well information retrieval, access to “Linked Open Data” (LOD) and data integration.

Ontology-based Clustering of Biological Linked Open Data

Increasing amounts of biomedical data provided as Linked Open Data (LOD) offer novel opportunities for knowledge discovery in bio-medicine. We proposed an approach for selecting, integrating, and mining LOD with the goal of discovering genes responsible for a disease [99] . We are currently working on the integration of LOD about known phenotypes and genes responsible for diseases along with relevant bio-ontologies. We are also defining a corpus-based semantic distance. One possible application of this work is to build and compare possible diseaseomes, i.e. global graphs representing all diseases connected according to their pairwise similarity values.

Suggesting Valid Pharmacogenes by Mining Linked Open Data and Electronic Health Records

A standard task in pharmacogenomics research is identifying genes that may be involved in drug response variability and called “pharmacogenes”. As genomic experiments in this domain tend to generate many false positives, computational approaches based on background knowledge may generate more valuable results. Until now, the later have used only molecular networks databases or biomedical literature. We are studying and working on a novel method that take advantage of an eclectic set of linked data sources to validate uncertain drug–gene relationships, i.e. pharmacogenes [3] . One advantage relies on the standard implementation of linked data that facilitates the joint use of various sources and makes easier the consideration of features of various origins. Accordingly, we proposed an initial selection of linked data sources relevant to pharmacogenomics. We formatted these data to train a random forest algorithm, producing a model that classify drug–gene pairs as related or not, thus validating candidate pharmacogenes.

With this same motivation of validating state-of-the-art knowledge in pharmacogenomics, a new ANR project called “PractiKPharma” will be initiated in 2016 and will rely on similar ideas. The originality of “PractiKPharma” is to use “Electronic Health Records” to constitute cohorts of patients that are then mined for validating extracted pharmacogenomics knowledge units (http://practikpharma.loria.fr/ ).

Biological Data Aggregation for Knowledge Discovery

During this year, in collaboration with the Capsid Team, we contributed to write up two multi-disciplinary projects with a group of clinicians from the Regional University Hospital (CHU Nancy) and bio-statisticians from the Maths Lab (IECL). The first project, entitled ITM2P (“Innovations Technologiques, Modélisation et Médecine Personnalisée”) lying in the so-called CPER 2015–2020 framework, was accepted and granted. The funding is mainly intended for medical and computing equipments and will be used to set up four scientific platforms. We are involved in the SMEC platform as a support for “Simulation, Modeling and Knowledge Extraction from Bio-Medical Data”.

The second project is a RHU (“Recherche Hospitalo-Universitaire”) project entitled Fight Heart Failure (FHF) and was accepted as a so-called “investissement d'avenir” and granted. We are in charge of a workpackage which will give us the opportunity of exploring important research questions. Among these questions, one is to define “data aggregation” mechanisms with a twofold objective: (i) the definition of pairwise patient similarity given that patients are described by complex dimensions involving relations and time and (ii) the efficient clustering of patients based on this similarity measure. Each cluster should correspond to a bioprofile, i.e. a subgroup of patients sharing the same form of the disease and thus the same diagnosis and care strategy. For doing that, we are currently investigating consensus theories [95] and their applicability to a bio-medical context, and as well aggregation operators as defined in various contexts, e.g. databases, data-warehouses, web of data, and graph theory. The idea is to consider relational and temporal data aggregation as a first class citizen in the data preparation phase of the knowledge discovery. This allows to assess the contribution of aggregation for such a task and in this context.

Another question is related to the construction of a prediction model for each bioprofile/subgroup –once validated by the clinicians– to be used in a decision support system. This will likely require the combination of symbolic and numerical methods for the classification task.

Analysis of biomedical data annotated with ontologies

Annotating data with concepts of an ontology is a common practice in the biomedical domain. Resulting annotations define links between data and ontologies that are key for data exchange, data integration and data analysis. Since 2011, we collaborate with the National Center for Biomedical Ontologies (NCBO) to develop a large repository of annotations named the NCBO Resource Index. This repository contains annotations of 36 biomedical databases annotated with concepts of more than 200 ontologies of the BioPortal (http://bioportal.bioontology.org/ ). In the preceding years, we compared the annotations of a database of biomedical publications (Medline) with two databases of scientific funding (Crisp and ResearchCrossroads) to profile disease research. One main challenge is to mine these annotations.

As a first attempt, we adapted pattern structures to analyze the annotations of biomedical databases [85] . We considered annotated biomedical documents as objects and the corresponding annotations were classified according to various dimensions, i.e. a particular aspect of domain knowledge. The resulting classification of annotations allowed not only to discover correlations between annotations but also incomplete annotations that could be fixed afterward. This adaptation of pattern structures opens many perspectives in term of ontology reengineering and knowledge discovery.