In yub18/DESSERT: Differential Analysis of RNA Structure Probing Experiments at Nucleotide Resolution: Uncovering Regulatory Functions of RNA Structure

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%",
  warning = FALSE,
  message = FALSE
)

DiffScan

DiffScan is a computational framework for differential analysis of RNA structure probing experiments at nucleotide resolution. Specifically, it identifies RNA structurally variable regions (SVRs) between two cellular conditions.

DiffScan is compatible with various RNA structure probing platforms.

System Requirements

DiffScan is built and tested upon R (version 4.0.2). It mainly depends on the following R packages.

dplyr, rlist, pipeR, pbmcapply, MASS, coin, plyr, ggplot2, seqinr, SpatialExtremes, preprocessCore

Installation

Install DiffScan in R or Rstudio with:

if (!requireNamespace("devtools", quietly = TRUE))
    install.packages("devtools")
if (!requireNamespace("preprocessCore", quietly = TRUE))
    install.packages("http://bioconductor.org/packages/3.11/bioc/src/contrib/preprocessCore_1.50.0.tar.gz")

devtools::install_github("yub18/DiffScan")

Typical install time: a few minutes or less.

Usage

Suppose there are two RNA transcripts (t1, t2) with structure probing data in condition A and B, with 2 replicates in each condition (A1, A2, B1, B2). To identify SVRs in t1, t2, first format the data like the following.

library(DiffScan)
library(dplyr)
set.seed(123)
r = list(
  t1 = tibble(A1 = runif(50), A2 = A1 + runif(50, -0.1, 0.1),
              B1 = A1 + runif(50, -0.1, 0.1), B2 = B1 + runif(50, -0.1, 0.1)),
  t2 = tibble(A1 = runif(150), A2 = A1 + runif(150, -0.1, 0.1),
              B1 = c(A1[1:90] + runif(90, -0.1, 0.1), rep(0, 10), 
                     A1[101:130] + runif(30, -0.1, 0.1), rep(1, 20)), 
              B2 = B1 + runif(150, -0.1, 0.1))
)

In this example, transcript t1 has no SVRs, and transcript t2 has 2 SVRs covering nucleotide positions 91 nt ~ 100 nt and 131 nt ~ 150 nt.

Second, initialize DiffScan as follows. (See the document of DiffScan::init for appropriate quality control.)

r_init = init(r, qc_ctrl = F, ncores = 1)

Third, perform normalization to remove systematic bias in reactivities. (Users convinced of the comparability between reactivities can skip this step.)

r_norm = normalize(r_init, ncores = 1)

Next, run the Scan algorithm to identify SVRs.

res = scan(r_norm$r, seed = 123, alpha = 0.05, ncores = 1)

DiffScan predicts no SVRs for transcript t1, and predicts 2 SVRs for transcript t2 covering nucleotide positions 132 nt ~ 150 nt and 92 nt ~100 nt.

res

Expected run time: a few minutes or less.

Flu dataset

The following code reproduces DiffScan's results for the Flu dataset.

data(Flu)
r = list(Flu = Flu[,1:8])
r_init = init(r, ncores = 1)
r_norm = normalize(r_init, ncores = 1)
res = scan(r_norm$r, seed = 123, alpha = 0.05, ncores = 1)

Citation

Yu, et al. Differential Analysis of RNA Structure Probing Experiments at Nucleotide Resolution: Uncovering Regulatory Functions of RNA Structure. Nature Communications, 2022.

Scripts and data

The scripts and data for the reproduction of the analyses and results in DiffScan paper are available here.

yub18/DESSERT documentation built on June 2, 2025, 12:13 a.m.