David McColgin, Paul Hoover, and Mark Igra
Abstract —The DataSpace for HIV vaccine studies is a discovery tool available on the Web to hundreds of investigators. We designed it to help them better understand activity in the field and explore new ideas latent in completed research. In the past, only small central groups with specific goals have integrated the data from these studies. Further data integration and broader access were hindered by both experimental complexity and a research culture that treats data as proprietary. Responding to a call for change by leading vaccine organizations, the DataSpace harmonizes immunoassay results and study metadata so that a broader research community can pursue more flexible discovery than centrally planned analyses. Insights from human-centered design and beta evaluation suggest important differences from typical visual analytics needs that may inform similar efforts in open scholarship. The contribution of this paper is to elucidate some of these differences and demonstrate an application that addresses them. We made several changes to familiar visualizations to support key tasks such as identifying and filtering to a cohort of interest, making meaningful comparisons of time series data from multiple studies that have different plans, and preserving analytic context when making data transformations and comparisons that would normally exclude some data. Additional features provide the detailed metadata needed to plot and interpret others’ work. We also discuss beta evaluation and the path to achieving domain impact.
Index Terms —Hypothesis forming, Time series visualization, Visual knowledge discovery, Vaccines, Public health