In connorhknight/IBRAP: Integrated Benchmarking scRNA-seq Analytical Pipeline (IBRAP)

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

klippy::klippy(tooltip_message = 'Click to copy', tooltip_success = 'Done', position = c('bottom', 'right'))

RShiny Application Tutorial

Activating The App

In this tutorial we will explore how to use the rshiny application that create a GUI for users to observe results. We can activate the application using the following command:

run.IBRAP.rshiny()

knitr::include_graphics(path = 'app_1.png')

Uploading Results

Next, we must upload our resultant file. The IBRAP objects must have been saved using the following command:

saveRDS(object = ibrap.obj, file = 'ibrap.rds', compress = T)

to upload your file: - select 'browse...' and choose your file. - once the file is loaded in, select activate.

## Plotting Cell Assignments

knitr::include_graphics(path = 'app_3.png')

Now we can begin to explore the data. We can select between method assays on the left panel alongside the reduction technique contained within. On the right, we can create an interactive plot that will be based on the selected reduction technique. In the panel in the centre, we can select which dataframe of cluster results (or metadata) to access (this is constrained to relevant results to the selected reduction method), the point size is also adjustable below these selection boxes. Finally, we press plot and an interactive plot of the results is created. With this plot you are able to zoom in, remove certain clusters from the plot and even save it!

Interpreting Benchmarking Results

knitr::include_graphics(path = 'app_4.png')

Underneath this we have some explanatory information from the benchmarking that was produced (this is not essential). The first box contains cluster assignemnt benchmarking. This either comes with 3 (no ground truth) or 5 (with ground truth) metrices.

• ASW, Average Silhouette Width determines the separation of a cluster to its closest neighbour cluster

• Dunn Index evaluates the compactness of a cluster and its distance to its closest neighbour cluster

• connectivity determines how connecetd the cluster assignments points are to eachother

• ARI, Adjusted Rand Index measures the agreement between the cluster assignments between the ground truth and novel assignments, this metric is adjusted for randomness

• NMI, Normalised Mutual Information functions similarly to ARI however it is adjusted for cluster sizes

Underneath this we have a boxplot that will describe the ASW between batches if batch correction was performed. A higher value indicates higher batch effects whilst a lower demonstrates less.

Viewing Gene Information

Once we have finished assessing our clustering results we can begin to understand the biology that drove the clustering. For this, we can select the feature tab on the left panel. WARNING: this may take some time to load since there are lots of gene names to load!

Feature Plots

knitr::include_graphics(path = 'app_5.png')

In this first part of the page we can produce multiple feature plots to observe (within a limit that fits the box). We can even filter the cells down to a cenrtain percentile range. the percentage is in decimals, i.e. 0.5 = 50%

Violin Plots

knitr::include_graphics(path = 'app_6.png')

We can also use violin plots to observe the data in a more categorical fashion

Dot Plots

knitr::include_graphics(path = 'app_7.png')

If we wish to observe a high dimension of genes at once, we an also opt for dot plots.

We hope the RShiny application creates an easy-to-use platform to observe IBRAPs results. Please leave any recommendations of issues within the issue section of our github!

connorhknight/IBRAP documentation built on March 9, 2023, 7:01 p.m.