DABI
Data Archive BRAIN Initiative
An archive with new analysis tools to analyze electrophysiology data.
Client
University of Southern California
University of California, Los Angeles
Year
2018-2019
Inventing a new interface for electrophysiology data analysis
Analyzing electrophysiology (Ephys) data has often been referred to as the Wild West. Every lab creates its own system for analyzing their study’s data, making it difficult if not impossible to combine results for larger study. The DABI project’s goal was to create ways to pool disparate studies that previously couldn’t be compared and then enable researchers to run analyses. I designed the interface that helped overcome these barriers.
Challenges
Participation
Getting buy-in on the project from a wide range of electrophysiology researchers proved to be difficult. NIH funding requires data sharing for many of these kinds of studies, but their unspoken question was “what’s in it for us?” since they had already done their own analyses and in some cases published the results. It was difficult to know exactly what types of analyses we would ultimately be able to offer, and we couldn’t give them mockups of any design for them to envision it. Obtaining the data also took time and required programmers to provide a way for researchers to share it that was HIPAA compliant, so an interface was needed but without knowing what the later site would be.
Data disparities
To decide how the interface would function I spoke with researchers to determine steps they would take to run various analyses. Each was specialized to their specific study or lab. Overcoming the differences to find the similarities, harmonize the data, and find a workflow that was most useful would require giving users a lot of flexibility on how to view the data details, pool the data, and run analyses.
Approach
Researcher controls
Principle investigators of each study were given a way to provide us their data and given full control over access to it. A HIPAA compliant interface was created along with a dashboard for them to control their data visibility and to manage requests. They could give permissions for others to become administrators, set visibility dates, view data request emails and respond to them, and view histories of permissions.
Exploring data
The data provided by researchers showed we could create several flows. Variables between studies could vary widely, but we determined that we would create two distinct flows; researchers could begin by exploring data from an individual study or pooling specific studies, or they could select harmonized variables across multiple studies.
Exploring by study
For researchers interested in a specific study, a flow was created by allowing them to select each study and view the metadata and de-identified details, such as the number of subjects, genders, age range, and diagnoses. By visiting the study page users could view the study’s publications and could make requests directly to the principal investigator for data access. Selecting multiple studies allowed users to pool them for analysis, or to request study data from multiple studies simultaneously.
Exploring variables
The second flow allowed users to examine harmonized variables, combine them from all studies or from specific ones, and save them for analysis. They could also create their own variables. Once they created the cohorts they could select the analysis tab and select the type of analysis they wanted to run.
Analyses
We determined we would be able to give users the options to run a linear regression, a logistic regression, and a spectral analysis. Once selected the required fields would appear and they could choose Ephys feature extraction, such as alpha power and millisecond peaks. They could select their own saved cohorts or example cohorts we created.
Results
The interface has been praised as being intuitive and easy to use, and the flow for users has continued to expand. The number of researchers using the site has grown steadily and the grant was recently renewed for another five years. Partnerships have enabled deeper analysis tools to be incorporated, such as Rave and Jupyter, providing a robust platform for new discoveries in the field. The pre-login pages were designed explaining the site’s purpose and functions, providing additional information and ways for researchers to further participate.