In compbiomed/singleCellTK: Comprehensive and Interactive Analysis of Single Cell RNA-Seq Data

knitr::opts_chunk$set(warning = FALSE)

Introduction

singleCellTK offers a convenient way to compute and select the most variable features that show the highest biological variability to use them in the downstream analysis. Feature selection methods available with the toolkit include vst, mean.var.plot & dispersion from Seurat [@Butler2018][@Stuart2019][@Satija2017][@Hao2021] package and modelGeneVar from Scran [@Aaron2016] package. Users can additionally use visualization options and methods to visualize the top variable genes interactively as well as from R console.

To view detailed instructions on how to use these methods, please select 'Interactive Analysis' for using feature selection in shiny application or 'Console Analysis' for using these methods on R console from the tabs below:

````{=html}

## Workflow Guide

````{=html}
<div class="tab">
  <button class="tablinks" onclick="openTab(event, 'interactive')" id="ia-button">Interactive Analysis</button>
  <button class="tablinks" onclick="openTab(event, 'console')" id="console-button">Console Analysis</button>
</div>

<div id="interactive" class="tabcontent">

1. Select Feature Selection & Dimensionality Reduction tab from the top menu. This workflow guide assumes that the data as been previously uploaded, filtered and normalized before proceeding with this tab.

\

2. Select Feature Selection sub-tab (selected by default) to open up the feature selection user-interface.

\

3. The Feature Selection sub-tab is divided into three panels namely, a) Compute variability metric, b) Select number of variable features, and c) Plot.

\

The working of sections a, b and c are described below:

a) Compute variability metric

\

The Compute variability metric panel performs the calculation of the variability of all the features present in the dataset. Users need to select a method either from Seurat (vst, mean.var.plot, dispersion) or Scran (modelGeneVar) packages. A expression matrix must be selected for the calculation. Note that Seurat - vst method requires an raw count matrix, while other methods requires normalized expression. After selection, click on Run to start the calculation.

After computation, a scatter plot of all features will be shown in the Plot panel, where the axes are from the computed metrics. The variability metrics are stored as feature metadata variables, and could be used for ranking the features.

b) Select number of variable features

\

Once step (a) is done, subset of highly variable features can be selected for downstream analysis basing on available metrics. Users need to select a metric that has been calculated, enter the number of the top variable features to select, and name the subset. After confirming all the setting, click on Run to create the subset.

Upon the creation of a new subset, the Plot will be updated: the scatter plot will be based on the selected metric, all the selected highly variable features will be labeled in red, and a few top features will be highlighted with text labels.

c) Plot

\

From the dropdown panel at the top of Plot panel, users are allowed to:

• Update plot with different metrics, i.e. coordinates of points.
• Label features with different subsets. Note that a subset created from a certain method can be labeled on the coordinates from a different method, for comparison of the variability ranking.
• Adjust the number of top features to be highlighted with text labels.
• Change the feature ID type. This should be available in feature metadata. We recommend preset the feature display ID type when importing the data set, so that this setting applies globally through out the toolkit.

````{=html}