SNOWFLAKE (Since 2014)
Knowledge discovery from linked data and clinical notes*
*renewed since 2017 as SNOWBALL: Discovering knowledge on drug response variablity by mining electronic health records
- Dr. Adrien Coulet, ORPAILLEUR project-team, Nancy Grand Est
- Dr. Nigam Shah, Stanford Center for Biomedical Informatics Research, Stanford University
The aim of SNOWFLAKE is to study the impact of connecting Electronic Health Records (EHRs) with Linked Open Data (LOD), because the team believes that LOD and associated ontologies can guide the process of knowledge discovery from EHRs. Scientific objectives of SNOWFLAKE are:
- To increase the quality of information extraction from EHRs.
- To create accurate links between EHR extracted data and LOD.
- To discover knowledge that explains the variability of drug responses.
Depending on their genetics and their exposure to the environment, individuals are reacting differentially to drugs. During its first year, the SNOWFLAKE Associate Team proposed to characterize this variability to drug response by mining the EHRs of the Stanford Hospital. The team implemented an approach to detect dose and drug changes in patient history with the hypothesis that if a patient over reacts or under reacts to a drug, physicians will either change the dosage or change the drug itself. In the meanwhile, the team is profiling phenotypes before and after dose and drug changes to characterize these events. Next step is to establish how phenotype profiles are specific to some drugs or to certain populations, and further how they can be used to predict the necessity of changing the dose or the drug of a patient.
Publications and Awards:
- 3 Journal articles & 2 Conference papers.
- First common article in preparation.
- Best Application Prize, NCBO Hackathon 2014 – for “Whypothesis?”, a prototypical software that searches the LOD for molecular mechanisms that could explain drug side effects of unknown origin
Kevin Dalleau, Yassine Marzougui, Sébastien Da Silva, Patrice Ringot, Ndeye Ndiaye, Adrien Coulet: Learning from biomedical linked data to suggest valid pharmacogenes. Journal of Biomedical Semantics, BioMed Central, 2017, 75.
The joint research project was renewed in 2017 for 3 years as SNOWBALL: Discovering knowledge on drug response variablity by mining electronic health records. In 2017-18, Adrien Coulet will spend a sabbatical year in the Shah lab at Stanford.