bioconductor-mast

name: bioconductor-mastdescription: Methods and models for handling zero-inflated single cell assay data.when_to_use: "Use when: Zero-Inflated scRNA-seq DE: Performing differential expression analysis on single-cell RNA-seq data using a Hurdle model (zlm) to account for bimodal expression patterns.; Data Filtering: Filtering outlier cells and wells where discrete and continuous parts of the signal deviate significantly, visualized via plotSCAConcordance and applied via mast_filter.; Two-Sample Testing: Conducting combined n. Not for: Raw Integer Counts: For un-normalized, raw integer counts; use zinbwave instead because MAST expects log-transformed, approximately scale-normalized data.; Basic QC and Visualization: For simple quality control metric calculation without differential"user-invocable: false

MAST

Dependencies & Environment

Package-intrinsic requirements from the Bioconductor landing page — reproduce in any R environment.

• Version: 1.38.0 · Bioconductor: 3.23 · R: ≥ 4.6
• Depends: SingleCellExperiment
• Imports: Biobase, BiocGenerics, S4Vectors, data.table, ggplot2, plyr, stringr, abind, reshape2, SummarizedExperiment, progress, Matrix
• Install: BiocManager::install("MAST")

When to Use

• Zero-Inflated scRNA-seq DE: Performing differential expression analysis on single-cell RNA-seq data using a Hurdle model (zlm) to account for bimodal expression patterns.
• Data Filtering: Filtering outlier cells and wells where discrete and continuous parts of the signal deviate significantly, visualized via plotSCAConcordance and applied via mast_filter.
• Two-Sample Testing: Conducting combined normal theory/binomial tests between pairs of groups using the LRT function.

When NOT to Use

• Raw Integer Counts: For un-normalized, raw integer counts; use zinbwave instead because MAST expects log-transformed, approximately scale-normalized data.
• Basic QC and Visualization: For simple quality control metric calculation without differential expression; use scater instead because it provides dedicated functions for QC and PCA plotting.

Data Requirements

• Input Format: A SingleCellAssay object, which can be constructed from a data.frame (FromFlatDF), a matrix (FromMatrix), or upcast from a SingleCellExperiment (SceToSingleCellAssay).
• Normalization State: Log-transformed, scale-normalized data that has been thresholded (e.g., $\log_2(\text{TPM} + 1)$).
• Structure: Supports dense matrices, sparse Matrix objects, and HDF5Array backends for out-of-memory data.

Key Parameters

• sca: The SingleCellAssay object containing the expression data and metadata.
• formula: Symbolic notation specifying the covariates for the Hurdle model in zlm (e.g., ~ Population + Subject.ID).
• method: The modeling function wrapper used in zlm (e.g., 'glmer' for mixed models or 'glm').
• ebayes: Logical in zlm indicating whether to use an empirical Bayes adjustment for the dispersion estimate.
• useContinuousBayes: Logical in zlm to employ Bayesian linear regression for the continuous component.
• exprs_value: Character string indicating which assay slot contains the log-like data to operate on (e.g., 'logcounts').

Best Practices

• Upcast existing SingleCellExperiment objects to SingleCellAssay using SceToSingleCellAssay to ensure compatibility and validity checks.
• Explicitly name the assay slot containing log-like data (e.g., exprs_value = 'logcounts') when calling zlm if it is not the default.
• Enable multicore support to speed up model fitting by setting options(mc.cores=4) before running zlm.
• Use burdenOfFiltering and plotSCAConcordance to visualize the impact of filtering criteria before applying them.

Common Pitfalls

• Providing Raw Counts: Passing raw integer counts to zlm leads to suboptimal model performance; Fix: Ensure the input assay contains log-transformed, scale-normalized data (e.g., log2 TPM + 1).
• Memory Exhaustion: Running linear models on extremely large datasets exceeds RAM; Fix: Store the data using a sparse Matrix or HDF5Array backend and pass it to FromMatrix.
• Missing Assay Detection: MAST fails to find the correct expression values if the assay is uniquely named; Fix: Ensure the assay name matches MAST's default log-like names (e.g., 'et', 'logcounts') or specify it explicitly.

Alternatives

• zinbwave: Uses zero-inflated negative binomial models directly on integer count data rather than requiring log-transformed inputs.
• scater: Focuses on pre-processing, quality control, and visualization of scRNA-seq data rather than complex hurdle-model differential expression.
• SCnorm: Designed specifically for the robust normalization of single-cell RNA-seq data rather than statistical testing.

Citations

• Finak, Greg, Andrew McDavid, Masanao Yajima, Jingyuan Deng, Vivian Gersuk, Alex K Shalek, Chloe K Slichter, et al. 2015. “MAST: A Flexible Statistical Framework for Assessing Transcriptional Changes and Characterizing Heterogeneity in Single-Cell RNA Sequencing Data.” Genome Biol. 16 (1): 1–13.
• McDavid, Andrew, Lucas Dennis, Patrick Danaher, Greg Finak, Michael Krouse, Alice Wang, Philippa Webster, Joseph Beechem, and Raphael Gottardo. 2014. “Modeling Bi-Modality Improves Characterization of Cell Cycle on Gene Expression in Single Cells.” PLoS Comput. Biol. 10 (7): e1003696.

References

• Homepage: https://bioconductor.org/packages/MAST
• Vignette: https://bioconductor.org/packages/release/bioc/vignettes/MAST/inst/doc/MAST-Intro.html

---

Run this on BioMate

This skill is the knowledge layer — when, why, and how to use mast. To run this analysis on your own data with managed compute, automated QC, and reproducible outputs, use BioMate — free to start.

▶ Open mast on BioMate →