A flexible denormalization technique for data analysis above a deeply-structured relational database: biomedical applications
| Authors | |
|---|---|
| Year of publication | 2015 |
| Type | Article in Proceedings |
| Conference | Lecture Notes in Computer Science 9043, Bioinformatics and Biomedical Engineering, Third International Conference, IWBBIO 2015, Granada, Spain, April 15-17 2015, Proceedings, Part I |
| MU Faculty or unit | |
| Citation | |
| Web | http://link.springer.com/chapter/10.1007%2F978-3-319-16483-0_12 |
| Doi | http://dx.doi.org/10.1007/978-3-319-16483-0_12 |
| Field | Informatics |
| Keywords | relational database; PostgreSQL; NoSQL; data flattening; automatic data denormalization |
| Description | Relational databases are sometimes used to store biomedical and patient data in large clinical or international projects. This data is inherently deeply structured, records for individual patients contain varying number of variables. When ad-hoc access to data subsets is needed, standard database access tools do not allow for rapid command prototyping and variable selection to create flat data tables. In the context of Thalamoss, an international research project on beta-thalassemia, we developed and experimented with an interactive variable selection method addressing these needs. Our newly-developed Python library sqlAutoDenorm.py automatically generates SQL commands to denormalize a subset of database tables and their relevant records, effectively generating a flat table from arbitrarily structured data. The denormalization process can be controlled by a small number of user-tunable parameters. Python and R/Bioconductor are used for any subsequent data processing steps, including visualization, and Weka is used for machine-learning above the generated data. |
| Related projects: |