The Signature Regulatory Clustering (SiRCle) method
integrates DNA methylation, RNA-seq and proteomics data at the gene
level to deconvolute the association between dysregulation within and
across possible regulatory layers (DNA methylation, transcription and/or
translation). Based on logical regulatory rules, sircleRCM, genes are
grouped in SiRCle clusters based on the layer (DNA methylation,
transcription and/or translation) where dys-regulation first occurs.
Using the output of sircleRCM, the SiRCle clusters, one can find the
primary biological processes altered by applying Over Representation
Analysis (ORA) (sircleORA) and the drivers behind it using
Transcription Factor (TF) analysis (sircleTF).
Lastly, to compare patient’s subsets (e.g. based on stage), we found
that integrating across the data layers prior to performing differential
analysis and biological enrichment better captures the biological
signal. Hence we use a variational autoencoder (VAE) to learn gene-wise
relationships across the three data layers to obtain an integrated value
for each gene (sircleVAE). Unsing the integarted value we next
perform a Mann-Whitney U test to identify genes with a significant
integrated difference between the patient’s groups.
sircleRCM functions create the SiRCle Regulatory Clustering Model
(RCM) based on logical regulatory rules, which in turn can be used for
further downstream analysis:
- Over Representation Analysis (
sircleORA) on the individual SiRCle clusters using GO-term pathways or your own pathway list of choice, which returns the results DFs and Visualisations of the results - Transcription Factor (
sircleTF) analysis using over representation analysis (ORA) on the individual SiRCle clusters using amy TF regulon input (e.g. dorothea or FIMO) and returns the results and TF visualisation - Variational Autoencoder statistics currently only runs with the python
package
See R tutorial in the vignettes folder, which includes a tutorial
and data to try the functions.
If you want to read more about how SiRCle works, please check out our
preprint
SiRCleR is an R package. 1. Install Rtools if you haven’t done
this yet, using the appropriate version
(e.g.windows or
macOS). 2. Install the
latest development version from GitHub with: SiRCleR package
direcly in R:
devtools::install_github("https://github.com/ArianeMora/SiRCleR") library(SiRCleR)
### Dependencies If you are using the visualisations for the over
representation analysis you will need to install the following tools and
cite them.
-
CRAN packages
install.packages('ggnewscale')
-
Biocmanager packages
if (!require("BiocManager", quietly = TRUE))
install.packages("BiocManager")
BiocManager::install("enrichplot")
BiocManager::install("org.Hs.eg.db")
BiocManager::install("clusterProfiler")
While we have done our best to ensure all the dependencies are documented, if they aren’t please let us know and we will try to resolve them.
Note if you are running Windows you might have an issue with long
paths, which you can resolve in the registry on Windows 10:
Computer Configuration > Administrative Templates > System > Filesystem > Enable Win32 long paths
(If you have a different version of Windows, just google “Long paths
fix” and your Windows version)
See the vignettes folder for a proper tutorial including data you can run the functions with. FYI: currently under development!
Please post questions and issues related to SiRCleR functions on
the Issues section of this GitHub repository.
If you want to reproduce the results of of our publication, please use the python package version found here: https://doi.org/10.1101/2022.07.02.498058
Ariane Mora, Christina Schmidt, Brad Balderson, Christian Frezza & Mikael Bodén 2022. SiRCle (Signature Regulatory Clustering) model integration reveals mechanisms of phenotype regulation in renal cancer. BioRxiv. https://doi.org/10.1101/2022.07.02.498058