The acorde R package contains an implementation of the pipeline showcased in Arzalluz-Luque et al. 2021 [1]. acorde is an end-to-end pipeline designed for the study of isoform co-usage networks using single-cell RNA-seq data (scRNA-seq).
The pipeline includes three basic analysis blocks:
-
Single-cell isoform quantification and Differential Expression (DE) filtering.
-
Detection of isoform co-expression relationships using percentile correlations and semi-automated clustering.
-
Differential and co-Differential Isoform Usage analysis. To couple these analysis with a biologically interpretable readout, we incorporate functional annotations onto isoform models, and use tappAS for functional analysis.
Since both the long read-transcriptome definition/quantification procedure and the functional analyses in [1] are based on external tools, the present R package does not include them.
To reproduce our pipeline's long read transcriptome building strategy, please refer to our manuscript's dedicated Supplementary Note [1].
Regarding functional analyses, we have included a section in the vignette providing instructions to perform them, including:
- How to obtain a functionally-annotated transcriptome using isoAnnotLite.
- How to generate input files that are compatible with the tappAS application.
acorde contains the necessary functions and documentation to obtain a set of DIU and co-DIU genes using an single-cell, isoform-level expression matrix as input. In addition, we provide instructions to reproduce the figures and additional analyses included in Arzalluz-Luque et al. [1]. The isoform expression matrix employed during the study is provided as internal data in the package.
Some of the co-expression metrics available in acorde are based on functions from the dismay R package. Since dismay is not available on CRAN, we suggest to install it from GitHub before starting the acorde installation. This can be done as follows:
install.packages("devtools")
devtools::install_github("skinnider/dismay")
If installing dismay requires propr, it should also be installed from GitHub.
devtools::install_github("tpq/propr")
The acorde R package and all the remaining dependencies can be installed
from GitHub using devtools:
devtools::install_github("ConesaLab/acorde")
To access vignettes, you will need to force building with
devtools::install_github(build_vignettes = TRUE). Please note that this will
also install all suggested packages required for vignette build and might
increase install time. Alternatively, an HTML version of the vignette is
available under the vignettes
folder.
In order to use acorde, you will need the following items:
- An isoform-level quantification matrix.
- Isoform-to-gene relationships and cell-level identity labels (i.e. cell type or state).
- A functionally-annotated transcriptome (provided that you wish to perform downstream functional analyses).
Please note that an example dataset (the one used in the acorde manuscript [1]) including all required objects is provided along with the package. Details to load and use them can be found in the package's vignette.
Importantly, acorde can be applied on both standard reference and long read-defined transcriptomes -the only limitation concerns the annotation of functional features for those isoforms.
For our study, we used IsoAnnotLite (v2.6) to transfer functional labels from tappAS’ pre-annotated mouse RefSeq 78 reference as well as from the mouse neural transcriptome used in the tappAS publication, which was possible thanks to the fact that we used a compatible reference for long-read processing (i.e. an updated version of the mouse RefSeq transcriptome). Therefore, users will need to take this into consideration and rely on pre-annotated references and isoAnnotLite until a de novo annotation tool is available (currently a work in progress).
In the meantime, all available references can be viewed and downloaded here. To use the mouse brain
transcriptome in our manuscript, download the
Mus_Musculus_GRCm38.p6_PacBioENCODE_RefSeq108.gff3 file.
If you encounter a problem, please open an issue via GitHub or send an email to angeles.arzalluz [at] gmail.com.
If you use acorde in your research, please cite the original publication:
[1] Arzalluz-Luque, A., Salguero, P., Tarazona, S. et al. acorde unravels functionally interpretable networks of isoform co-usage from single cell data. Nat Commun 13, 1828 (2022). https://doi.org/10.1038/s41467-022-29497-w