In “big data” analysis, we usually construct “analysis funnels”: In several steps, a huge amount of raw data gets reduced to medium-sized immediate data and a small number of final results. How do we know that our funnel works as intended? To check such an analysis, we need to be able to trace back any result to the intermediate or raw data that it summarizes. To do so visually, it is most helpful to “link” plots: Clicking on a data point in one plot displays the data underlying it in another plot, allowing one to “walk backwards” through the analysis pipeline.
We present R/LinkedCharts, a novel framework that allows bioinformaticians to construct interactive apps for any kind of application. With only very few lines of R code, one can set up “linked charts” such that clicking on data points in some plots changes what part of the data is shown in other plots.
I will demonstrate how this can be used to set up ad hoc visualization to walk through any kind of analysis, using as example two typical use cases: single-cell RNA-Seq and analysis of perturbation screens.
Participants who already know how to make ordinary static plots in R will learn how to make their plots interactive and “linked”, and how to use this for initial data exploration, pipeline development, hypothesis formation, quality assessment, and data dissemination. We will discuss not only R/LinkedCharts but also general strategy to perform powerful, sound, reliable and traceable big-data analyses.