An integrative pipeline for combining first and second generation cell type deconvolution results
Figure 1. A schematic overview of the `multideconv` pipeline
To avoid GitHub API rate limit issues during installation, we recommend setting up GitHub authentication by creating and storing a Personal Access Token (PAT). You can do this with the following steps:
# install.packages(c("usethis", "gitcreds"))
usethis::create_github_token() #Create a Personal Access Token (if you don't have)
gitcreds::gitcreds_set() #Add the tokenYou can install the development version of multideconv from
GitHub with:
# install.packages("pak")
pak::pkg_install("VeraPancaldiLab/multideconv")These are basic examples which shows you how to use multideconv for
different tasks. For a detailed tutorial, see Get
started
Before running multideconv, make sure to set your working directory.
The Results/ folder, where outputs will be saved, will be created in
this directory.
setwd('~/path/to/directory')
library(multideconv)The function calculates cell abundance based on cell type signatures
using different methods and signatures through the function
compute.deconvolution which takes as input the bulk RNAseq gene
expression matrix either as raw or normalized counts.
deconv = compute.deconvolution(raw.counts, normalized = T, credentials.mail = "xxxx", credentials.token = "xxxxxx", file_name = "Tutorial")
deconv = compute.deconvolution(raw.counts, normalized = T, credentials.mail = "xxxx", credentials.token = "xxxxxx", methods = c("Quantiseq", "MCP", "XCell", "DWLS"), file_name = "Test")
deconv = compute.deconvolution(raw.counts, normalized = T, credentials.mail = "xxxx", credentials.token = "xxxxxx", signatures_exclude = "BPRNACan", file_name = "Tutorial")
deconv = compute.deconvolution(raw.counts, normalized = T, credentials.mail = "xxxx", credentials.token = "xxxxxx", sc_deconv = T, sc_matrix = sc.object, cell_label = cell_labels, sample_label = bath_ids, name_sc_signature = "Signature_test", file_name = "Test")NOTE: CIBERSORTx is included in the deconvolution methods, but
it’s not an open-source program. To run it, please ask for a token in
CIBERSORTx and once
obtained, provided your username and password on the parameters
credentials.mail and credentials.token.
Also CIBERSORTx ask to have docker install in your computer. Once installed, be sure to concede permission to manage docker as a non-root user. For verify this, go to the terminal and run:
docker ps
If you don’t have any error, congrats you are go to go!
If you receive an error of permission, just run:
sudo groupadd docker
sudo usermod -aG docker ${USER}
Then restart your computer and try again
docker ps
Now, you should be able to run deconvolution using CIBERSORTx without any problems :)
For processing the deconvolution features obtained from
compute.deconvolution, you can use the
compute.deconvolution.analysis function.
processed_deconvolution = compute.deconvolution.analysis(deconvolution, corr = 0.7, seed = 123, return = T)Users can also compute second-generation deconvolution methods using
their single cell data. For this use the function
compute_sc_deconvolution_methods. Remember that this function is
already included in compute.deconvolution when setting
sc_deconv = T.
deconv_sc = compute_sc_deconvolution_methods(raw_counts, sc_object, sc_metadata, cell_annotations, samples_ids, name_object, normalized = T, n_cores = 4, cbsx_name = "XXX", cbsx_token = "XXX")multideconv works based on established cell naming conventions (Figure
2) to simplify analysis and processing. Thus, if you would like to use
your own deconvolution results or signatures, please make sure to follow
these formats.
Figure 2. Cell types nomenclature for `multideconv`
If you want multideconv to consider other cells, it is pretty simple!
Just use the argument cells_extra in the function
compute.deconvolution.analysis().
Let’s say you want to add mesenchymal and basophils cells:
processed_deconvolution = compute.deconvolution.analysis(deconvolution, corr = 0.7, seed = 123, cells_extra = c("mesenchymal", "basophils")) And that’s it, just make sure the name you are putting in cells_extra is exactly the name of your cells in your deconvolution matrix!
You can include other signatures into the analysis by adding them as
.txt into the folder Results/custom_signatures.
multideconv includes second generation methods that allows users to
input single cell data and use it for constructing signatures
‘on-the-fly’ and deconvolve the bulk RNAseq data.
Because most of the times scRNAseq data can be big and sparse matrices,
we applied ‘preprocessing’ steps to avoid crashing the computer. For
this, multideconv construct metacells per cell type and patient from
the single cell data using the KNN algorithm. This is done essentially
by executing different tasks per cluster in parallel for N workers and
once finished, it computes deconvolution using the reduced single cell
object.
NOTE: If you encounter missing packages for certain deconvolution
methods (e.g., MuSiC, SCDC) during the pipeline, make sure to
install the versions maintained by the omnideconv team. These versions
are patched to work seamlessly with omnideconv functions:
devtools::install_github("omnideconv/MuSiC")
devtools::install_github("omnideconv/SCDC")If you encounter any problems or have questions about the package, we encourage you to open an issue here. We’ll do our best to assist you!
multideconv was developed by Marcelo
Hurtado in supervision of Vera
Pancaldi and is part of the
Pancaldi team. Currently, Marcelo
is the primary maintainer of this package.
If you use multideconv in a scientific publication, we would
appreciate citation to the :
Hurtado, M., Essabbar, A., Khajavi, L., & Pancaldi, V. (2025). multideconv – Integrative pipeline for cell type deconvolution from bulk RNAseq using first and second generation methods. bioRxiv. https://doi.org/10.1101/2025.04.29.651220