Tutorial Registration Closed: Due to high demand, the tutorial sessions have reached seating capacity.
Monday, March 23
8:30 a.m. – 12:00 p.m.
T01: Authoring, Sharing, and Executing Electronic Phenotyping Algorithms: Past, Present and the Future
J. Pathak, Mayo Clinic; J. Denny, Vanderbilt University School of Medicine; L. Rasmussen, Northwestern University; W. Thompson, NorthShore University HealthSystem
This tutorial will cover basic themes about use of Electronic Health Records (EHRs) for generating cohorts of patients to serve as cases and controls for given clinical phenotypes. EHRs can be used for many different types of research include disease-based, response to treatment, clinical biomarkers, redefining “normal”, and analysis of changes over time of clinical variables and parameters. Deriving such phenotypes from EHR data can be challenging. Methods typically involve use of structured data such as billing codes, medication records, and laboratory data, as well as unstructured data such as narrative reports, for which natural language processing (NLP) system are often used. After derivation of these phenotypes, populations can be used for clinical research. Linkage to DNA biobanks also enables the possibility of genomic and pharmacogenomic associations. Research within the electronic MEdical Records and GEnomics (eMERGE; http://gwas.org) network has demonstrated success with EHR-based phenotypes for use in genome-wide association studies (GWAS). In addition, use of EHR-linked genetic data uniquely enables phenome-wide associations studies (PheWAS), which allows an unbiased scan of what diseases may be associated with a given genotype. This tutorial will review the design of EHR-linked biobanks; methods for creating phenotype algorithms with integrated use of NLP, billing codes, laboratory and test results, and medication records (with review of a number of case studies); use of standard vocabularies in representing phenotype data, use of standards-based approaches for constructing and representing phenotype algorithms; and application of PheWAS to further characterize clinical variants. The tutorial will also include a hands-on session using a novel platform for authoring, sharing and execution of electronic phenotyping algorithms (http://informatics.mayo.edu/PhEMA).
T02: Taking Control of Your Findings: A Guide to Experimental Approaches for Validating Mined Data
H. Fan Minogue, Stanford University School of Medicine
In the last decade, advances in the technology has drastically speeded up the harvest of biological data and generated unprecedented amount of information. Modern biology has become a data-driven science, which also posed considerable challenges to experimental scientists in data processing and extraction of biological meanings. Bioinformatics has taken an increasingly vital role to meet the need for the effective management, analysis and integration of the data. New insights about diseases and their molecular basis have been identified through mining of massive-scale datasets and some have even led to new therapies. However, the role of bioinformatics researchers is often defined as a service provider, who facilitates the discovery of new biological insights by providing new clues, yet the validation of their predictions often needs to be performed by experimental scientists. Thus bioinformatics researchers often have to hand over their data and wish positive feedback or end up as the middle author of the work. This tutorial will attempt to demonstrate the capability of data scientists to take control of their in silico findings by revealing the principle and application of approaches to experimental validation, the resources for validation assays, and the strategies to apply them for the specific questions of interest. Attendees should expect to learn about the essential experimental approaches used for validation and what they measure. The application of each assay and the resources for performing them will be discussed. The strategies to carry out a successful validation will also be discussed. Finally, the tutorial will end with three use cases showing what kind of assays can be utilized for each specific question, a discussion of hurdles and the cycle of experimentation and evaluation, and a brainstorming session about how these approaches can in turn empower and motivate the data mining.