Omic Data Modelling for Information Retrieval

This study aims at designing a data model managing the vast majority of omic data types, dedicated to information retrieval on two dimensional levels, single patient and multi-patients. The exploitation of scientific health research data, in the search for new biomedical applications, is a promising challenge. Data integration generated by scientific research, particularly omics, with clinical data stored in the Electronic Health Record (EHR) can lead to significant progress in the development of new diagnostic tests and therapies as well as improve our understanding of complex genetic diseases and cancers. Currently, a few solutions already exist such as i2b2 or Transmart. However, they do not handle all main omic data types. Moreover, integrating omic analysis results in EHRs is today mandatory to help clinicians in decision making. This study proposes a generic omic data model dedicated to managing main omic data types (expression data, DNA-methylation, variants, etc.), omic data representation, and information retrieval. Integrating omics data within clinical data involves: (i) identify the different types of omic data, (ii) select relevant information in the context of integration with clinical data then (iii) design an effective omic data model. Four levels of data have been defined according to their level of interpretation. The second level representing interpreted data has been selected to build the data model. Various omic data types have been integrated into a database coupled to clinical data.