| name | bioconductor-metapod |
| description | Implements a variety of methods for combining p-values in differential analyses of genome-scale datasets. Functions can combine p-values across different tests in the same analysis (e.g., genomic windows in ChIP-seq, exons in RNA-seq) or fo |
| when_to_use | Use when: Genomic Window Aggregation: Combining p-values across adjacent genomic windows in ChIP-seq (e.g., inside csaw) using grouped functions like groupedSimes.; Multi-Batch Marker Detection: Merging statistics from multiple batches or pairwise comparisons during marker gene detection (e.g., inside scran) using parallel functions like parallelSimes.; Consensus Direction Summarization: Determining the ove. Not for: For simple multiple testing correction on a single flat list of independent p-values, use base R's p.adjust instead because metapod is designed for hierarchical or multi-test aggregation.; For combining raw effect sizes or variances directly across i |
| user-invocable | false |
metapod
Dependencies & Environment
Package-intrinsic requirements from the Bioconductor landing page — reproduce in any R environment.
- Version: 1.20.0 · Bioconductor: 3.23 · R: ≥ 4.6
- Imports: Rcpp
- System requirements: C++11
- Install:
BiocManager::install("metapod")
When to Use
- Genomic Window Aggregation: Combining p-values across adjacent genomic windows in ChIP-seq (e.g., inside
csaw) using grouped functions like groupedSimes.
- Multi-Batch Marker Detection: Merging statistics from multiple batches or pairwise comparisons during marker gene detection (e.g., inside
scran) using parallel functions like parallelSimes.
- Consensus Direction Summarization: Determining the overall direction of effect (up, down, or mixed) across multiple tests using
summarizeParallelDirection or summarizeGroupedDirection.
When NOT to Use
- For simple multiple testing correction on a single flat list of independent p-values, use base R's
p.adjust instead because metapod is designed for hierarchical or multi-test aggregation.
- For combining raw effect sizes or variances directly across independent studies, use metafor instead because
metapod operates strictly on p-values and their directions.
Data Requirements
- Input Format: A numeric vector of p-values with a grouping factor (for
grouped* functions) or a list of parallel numeric vectors of p-values (for parallel* functions).
- Normalization State: P-values must be properly calibrated (uniform under the null hypothesis).
- Minimum Size: At least 2 p-values per group or parallel list to combine.
Key Parameters
- method ("simes"): The method used to combine p-values in wrapper functions like
combineGroupedPValues (e.g., "simes", "holm-min").
- log.p (FALSE): Logical indicating whether the input p-values are log-transformed and whether to return log-transformed output.
- influential: A logical vector (returned by combining functions) passed to
summarizeParallelDirection to only consider tests that contributed to the final p-value.
- weights: Numeric vector of weights to apply to individual tests (supported by Simes, Stouffer, and Holm-min methods).
Best Practices
- Use Simes' method (
parallelSimes or groupedSimes) when you expect dependencies between tests (e.g., overlapping genomic windows) as it is robust to such dependencies.
- Set
log.p = TRUE when dealing with extremely small p-values to prevent underflow to zero during calculation.
- Use
summarizeParallelDirection or summarizeGroupedDirection on the influential tests to avoid noise from non-significant tests when determining the consensus log-fold change direction.
Common Pitfalls
- Assuming Independence Incorrectly: Using Fisher's method (
groupedFisher or parallelFisher) on highly correlated tests. Fix: Switch to Simes' method or the minimum Holm approach (groupedHolmMin), which do not require independence.
- Underflow of P-values: Generating zero p-values due to numerical limits when combining highly significant tests. Fix: Pass log-transformed p-values and set
log.p = TRUE.
- Misinterpreting Direction Counts: Using
countParallelDirection without understanding that it applies Benjamini-Hochberg or Holm corrections internally. Fix: Read the documentation for countParallelDirection carefully, or use summarizeParallelDirection for a simpler consensus.
Alternatives
- poolr: For combining p-values while accounting for correlation using various methods.
- survcomp: Contains classical p-value combining methods (Fisher, Stouffer) but lacks genomic-specific optimizations and grouped/parallel vector handling.
- metafor: For traditional meta-analysis combining effect sizes and standard errors rather than just p-values.
Citations
- Lun, A. T. L. (2021). metapod: Meta-Analyses on P-Values of Differential Analyses. R package.
References
Run this on BioMate
This skill is the knowledge layer — when, why, and how to use metapod. To run this analysis on your own data with managed compute, automated QC, and reproducible outputs, use BioMate — free to start.
▶ Open metapod on BioMate →
