MARVEL is an R package developed for alternative splicing analysis at single-cell resolution. MARVEL complements published single-cell splicing softwares with the following features:

1. Percent spliced-in (PSI) quantification for all seven main exon-level splicing events, i.e. skipped-exon (SE), mutually-exclusive exons (MXE), retained-intron (RI), alternative 5' and 3' splice sites (A5SS, A3SS), and alternative first and last exons (AFE, ALE).
2. Stratify PSI distribution for each splicing event into the modalities (discrete splicing patterns), and adjust for technical biases during this assignment.
3. Integrated differential splicing and gene expression analysis to reveal gene-splicing dynamics.
4. Dimension reduction analysis.
5. Pathway enrichment analysis.
6. Splicing-associated nonsense-mediated decay (NMD) prediction.
7. Multiple visualisation functions for exploring splicing and gene expression across cell populations.
8. Supports both plate-based (e.g., Smart-seq2) and droplet-based (e.g., 10x Genomics) single-cell RNA-sequencing data analysis.
9. In principle, also applicable to bulk RNA-sequencing data analysis.

Please install the following pre-requisite R packages from CRAN prior to installing MARVEL.

install.packages("ggplot2")
install.packages("Matrix")
install.packages("plyr")
install.packages("scales")

MARVEL is available on CRAN.

install.packages("MARVEL")
library(MARVEL)

Alternatively, MARVEL may be installed from Github, which includes several functionalities in beta-testing phase.

library(devtools)
install_github("wenweixiong/MARVEL")
library(MARVEL)

The following packages are not mandatory for MARVEL installation, but are highly recommended to support the functionalities of MARVEL.

if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("AnnotationDbi")
BiocManager::install("Biostrings")
BiocManager::install("BSgenome")
BiocManager::install("BSgenome.Hsapiens.NCBI.GRCh38")
BiocManager::install("clusterProfiler")
BiocManager::install("GenomicRanges")
BiocManager::install("IRanges")
BiocManager::install("org.Hs.eg.db")
BiocManager::install("org.Mm.eg.db")
BiocManager::install("phastCons100way.UCSC.hg38")

The following packages are not mandatory for MARVEL installation, but are highly recommended to support the functionalities of MARVEL.

install.packages("factoextra")
install.packages("FactoMineR")
install.packages("fitdistrplus")
install.packages("ggplot2")
install.packages("ggrepel")
install.packages("gtools")
install.packages("kSamples")
install.packages("pheatmap")
install.packages("reshape2")
install.packages("S4Vectors")
install.packages("scales")
install.packages("stringr")
install.packages("textclean")
install.packages("twosamples")

Please install the modified wiggleplotr R package from here: https://drive.google.com/file/d/1eEkq1axQXynVifWanh9ACichIrJjyNfr/view?usp=share_link

install.packages("wiggleplotr_1.18.0.tar.gz", repos=NULL, type="source")

Single-cell plate-based alternative splicing analysis: https://wenweixiong.github.io/MARVEL_Plate.html
Single-cell droplet-based alternative splicing analysis: https://wenweixiong.github.io/MARVEL_Droplet.html

version 2.0.0

• First version uploaded

version 2.0.1

• Updated AnnotateGenes.10x function to enable handling of GTF with either gene_type or gene_biotype attribute label. Previously, only handled GTF with gene_biotype attribute label. This function is only applicable for droplet-based data.
• Updated ValidateSJ.10x function to enable filtering in of novel splice junctions. A novel splice junction is defined as one end mapping to a known exon while the other end mapping to a novel/unknown exon. Simply include the argument keep.novel.sj=TRUE (default is FALSE) to retain novel splice junctions for downsteam analysis. Novel splice junctions may be of specifically expressed in a particular disease or hitherto unreported splice junctions expressed in a particular tissue or cell type. Previously, only splice junctions with both ends mapping to known exons were retained while novel splice junctions were filtered out. This function is only applicable for droplet-based data.
• Included SubsetCrypticSS function to enable filtering in of cryptic A5SS and A3SS for the list of pre-defined A5SS and A3SS splicing events provided by the user (e.g., rMATS). A cryptic A5SS or A3SS is defined as novel splice site located within 100bp of the canonical splice site. Simply specify EventType=="A5SS" and then execute the same function with EventType=="A3SS" argument. Distance between novel and canonical splice sites may be specified using the DistanceToCanonical argument (default is 100). This function should be executed after creating the MARVEL object with the CreateMarvelObject function. This function is only applicable for plate-based data.
• Included RemoveCrypticSS function to enable filtering out of cryptic A5SS and A3SS for the list of AFE and ALE splicing events detected by MARVEL, respectively. Simply specify EventType=="AFE" and then execute the same function with EventType=="ALE" argument. This function should be executed after detecting AFE and ALE using the DetectEvents function. This function is only applicable for plate-based data.

version 2.0.2

• Updated AnnotateSJ.10x function to enable handling of GTF with chr prefix. Previously, only handled GTF without chr prefix. This function is only applicable for droplet-based data.

version 2.0.3

• Updated RunPCA function to enable a Bayesian approach for imputing missing PSI values prior to dimension reduction with principal component analysis (PCA). To use this approach for PCA, simply specify method.impute="Bayesian" and level="splicing". This approach was inspired by Benegas et al. (https://elifesciences.org/articles/73520) and please cite his work if you use this option. Note that the ComputePSI.Posterior function would need to be executed before running the RunPCA function. Currently, missing PSI values are imputed with random values between 0-100 with the method.impute="random". This function is only applicable for plate-based data.
• Updated CompareValues function to enable setting the random number generator when using DTS for differential splicing analysis. This will ensure reproducibility of differential splicing results. Simply specify seed.dts=1 (or any numeric value) together with method="dts". Currently, and by default, the seed is not set when DTS is used. This function is only applicable for plate-based data.

We are keen to further improve MARVEL to make it more comprehensive for single-cell splicing analysis. In particular we hope to include more functionalites related to functional annotation, e.g., predicting the biological consequence of alternative splicing. If interested please get in touch :)

Sean Wen [email protected].