In hyf97ca/hotgenes: Simulates Heat Physics on your RNA-seq Dataset for More Compelling Visualization

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

library(hotgenes)

Getting started

Running the following in sequence should get you an output plot:

    sm <- generateStrandModel(startBase=1, endBase=195471971,
     fc=musCh1fc, chr="chr1", strand="-", scaling=100000)
    lm <- generateLocationModel(startBase=1, endBase=195471971, scaling=100000)
    smh <- simulateHeatSpread(.strandModel=sm, conductivity=0.01, iterations=5)
    plotHeatedMap(smh, lm)

Explanation

The above code loads the example dataset, musCh1fc, which is included with the package. It is the output from featureCounts() from Rsubread. To load your own custom dataset, you can substitute your own data object in the fc parameter.

The central data structure of the hotgenes package is the strand model, which is a matrix of row vectors. Each row corresponds to one set of RNA-seq data from your featurecounts file. A strand model is made with a specific start base, end base, and scale. The first cell in the row represents aggregate RNA-seq readings from the start base (1) to start + scaling (100000) bases. the last row similarly represents the last scaling bases until the end base (195471971).

If you are using your own dataset, use relevant values for it; if your strand is three million base pairs long then endBase should be three million. Similarly, larger DNA strands will be processed much faster if scaling is higher. Aim for 1000 cells in the row vector; This will ensure a relatively fast calculation without sacrificing much resolution.

After the strand model is generated, it is (optionally) further processed by simulateHeatSpread(). This function models heat flow between cells in the strand model; 'heat' flows from high expression areas to low expression areas adjacent. The total 'heat' is unchanged. Essentially, individual cells are made less 'hot' but adjacent cells will share some of the heat, making the expression hotspot more visible while creating a glowing visual effect.

Finally, a plot is generated with plotHeatedMap(). This takes in a strand model (processed or not) as well as a row vector (location model) which maps each cell to where it should go across the DNA strand.

generateLocationModel() is a helper function that creates such a row vector based on the parameters given which should match that of generateStrandModel. Essentially it's a regular sequence generator that helps label the strand model.

Optionally a palette can be specified as per hcl.colors from the grDevices package. You should pick a palette recommended for heatmaps.

Shiny app

The shiny app can be started by running the following code:

runHotgenes()

This shiny app is a wrapper for the above functions with some additional optimizations.

Use your own data with the shiny app

You will have to download the package source and edit the shiny script.

To use your own data, edit the following code in app.R in inst/shiny-scripts:

#DNA strand to be rendered start bp
dnaStart <- 1
#DNA strand to be rendered end bp
dnaEnd <- 195471971
#1 is most flexible, might speed up if use 1000; the scaling options must be divisible by this
baseScale <- 1000
#chromosome number (required for reading featureCounts)
dnaChr <- "chr1"
#feature counts data structure to use
featureCounts <- musCh1fc

Specifically, point your own data to featureCounts, and edit dnaStart, dnaEnd, dnaChr as appropriate for that dataSet.

Limitations

• hotgenes is only capable of rendering one chromosome at a time. Since the featureCounts data structure is used for all chromosomes, the specific chromosome will have to be specified.

• hotgenes can render multiple datasets (for one chromosome) in the same plot because the featureCounts data structure supports multiple datasets. However, your computer's processing power (as of 2019) is probably too low for a snappy result; do so at your own discretion.

hyf97ca/hotgenes documentation built on Dec. 8, 2019, 12:41 p.m.