dStruct is now being maintained and supported by Aviran lab alumnus, Krishna Choudhary. Please visit https://github.com/dataMaster-Kris/dStruct for updates and to report issues. This repository contains the code for the initial release of dStruct reported in our publication.
Method for identifying differential reactive regions from RNA structurome profiling data.
dStruct is a statistical method for identifying regions that display altered reactivity patterns between two groups of samples. It can perform de novo discovery or identify regions from a list provided by the user. The latter case is called guided discovery. dStruct is compatible with a diverse range of structure profiling technologies, accounts for biological variation and controls the false discovery rate in a multiple testing context.
dStruct is written in R version 3.4.1. To start with, download and install the latest versions of R and RStudio.
dStruct can be installed directly from source. First, install devtools by executing the following in RStudio.
install.packages("devtools")
Next, in RStudio run the following.
library(devtools)
install_github("AviranLab/dStruct")
This should install the package in R. Check by executing the following command.
library(dStruct)
In the future, we plan to make dStruct available through Bioconductor.
dStruct takes reactivities of groups of samples, say A and B, for all transcripts under consideration. The reactivities for each transcript must be a data frame object with columns labeled as A1, A2, ... and B1, B2, ... The numerals in column names indicate sample number. Let us say that reactivity represents one such data frame, that there are 3 samples of each group and that the user needs to search for a minimum length of 11 nt. Note that this is the default value for this parameter. For other parameters that can be specified, refer the manual. Unless specified, dStruct assumes that the user intends to run the analysis with default settings. De novo discovery of differential regions can be done by executing the following command.
dStruct(reactivity, reps_A = 3, reps_B = 3, min_length = 11)
Reactivities for all transcripts can be stored together in a list object, say rlist, with one data frame for each transcript. All the list elements must have unique names. In this case, de novo discovery can be done simultaneously for all transcripts.
dStructome(rlist, reps_A = 3, reps_B = 3, min_length = 11)
Users can specify the number of cores to be used by dStructome for parallel processing of transcripts.
For guided discovery, data frame with reactivities should contain only information for regions of interest. There should be a separate data frame for each region. Let one such data frame be reac_region1. If there are say, 3 samples of each group, guided discovery can be performed as follows.
dStruct.guided(reac_region1, reps_A = 3, reps_B = 3)
All the data frames can also be combined in a single list object, say rlist. Then, guided discovery could be done in parallel for all regions using the following command.
dStructome(rlist, reps_A = 3, reps_B = 3, method = "guided")
For other options available in dStruct package, refer the dStruct manual or email the contributors.
Choudhary K., Lai Y.H., Tran E., Aviran S. (2019) "dStruct: identifying differentially reactive regions from RNA structurome profiling data."
- Krishna Choudhary - Initial implementation and developer
- Sharon Aviran - Supervisor