Skip to content

TOPMed Genotype-Phenotype Association Testing

Description

Scenario

Geraldine is a biomedical researcher who wants to look for genetic variants associated with coronary artery disease (CAD) by performing Genome Wide Association Analysis (GWAS), using genotypes from whole genome sequence (WGS). A large sample size is needed for statistical power to detect effects on a complex trait such as CAD. This sample size can be achieved only by harmonizing and combining data across multiple studies. WGS provides hundreds of millions of variants to test, so powerful computing resources are also required. Furthermore, Geraldine wants to collaborate on this project with a team of investigators who have expertise in various aspects of CAD, so needs a mechanism for sharing human subjects data.

Current Approach

Without the Commons, Geraldine would have to find and download several datasets to build a large synthetic cohort. It is likely that different studies would have annotated their data differently, making it difficult to identify data elements representing relevant aspects of CAD. Geraldine would need to spend a considerable amount of time searching data repositories and identifying appropriate variables because, for example, one study may name the relevant variable as CAD and another as atherosclerosis. Geraldine would also need to find a large computing resource to run the analysis and to install and run a number of command line programs. She would need to obtain permission to access the relevant data sets and download large data files containing the WGS data. Because the data are from human subjects, and require controlled access, they cannot readily be shared with collaborators.

With the Data Commons Phase I

Geraldine logs into her favorite full stack and searches for dbGaP studies with WGS-genotypes and CAD phenotypes. She is able to find CAD-related phenotypes in multiple studies, even though they use different variable names and descriptions, because the search process uses biomedical ontologies, natural language processing and other sophisticated tools. She obtains dbGaP approvals for these data sets (through the existing dbGaP mechanism) and is able to access the files within the Data Commons, eliminating the need for large file transfers. She has a workspace within which she can share data and results with her collaborators (who have the same access permissions). She can run her analysis using a pre-built and cost-efficient workflow. She and her collaborators can plot and summarize their results interactively within their workspace using a Jupyter notebook.

With Data Commons longer vision:

Geraldine now has the ability to search many more data repositories to find CAD-related studies. When she finds new studies, she is able to apply for access easily using a Data Commons interface, which prompts her for the necessary information. Her approvals arrive much faster than before due to automation of data access processes, and she is able guide her collaborators to efficiently obtain compatible approvals. Large-scale harmonization of the phenotype metadata has been done so that searches for CAD-related phenotypes have better sensitivity and specificity, thus facilitating the construction of her cross-study data set. She now has a large suite of pre-built applications and workflows to choose from for her analysis.

User Narratives

User narratives for this summary:

(Use Case Summary SUMMARY-1 - view source | edit)

Comments