bioconductor-bumphunter

name: bioconductor-bumphunterdescription: Tools for finding bumps in genomic datawhen_to_use: "Use when: Finding continuous, spatially clustered genomic regions (\"bumps\") that differ significantly between conditions using bumphunter().; Grouping genomic locations into distinct clusters based on maximum distance using clusterMaker().; Extracting positive, near-zero, and negative segments from a vector of test statistics using getSegments().; Packaging segmented regions into a table of bump characteris. Not for: For basic linear modeling of independent, unclustered genomic features, use limma because bumphunter is specifically designed to share information between nearby clustered locations.; For end-to-end analysis of Illumina 450k arrays without manual mat"user-invocable: false

bumphunter

Dependencies & Environment

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

• Version: 1.54.0 · Bioconductor: 3.23 · R: ≥ 4.6
• Depends: S4Vectors, IRanges, Seqinfo, GenomicRanges, foreach, iterators, locfit
• Imports: matrixStats, limma, doRNG, BiocGenerics, GenomicFeatures, AnnotationDbi
• Install: BiocManager::install("bumphunter")

When to Use

• Finding continuous, spatially clustered genomic regions ("bumps") that differ significantly between conditions using bumphunter().
• Grouping genomic locations into distinct clusters based on maximum distance using clusterMaker().
• Extracting positive, near-zero, and negative segments from a vector of test statistics using getSegments().
• Packaging segmented regions into a table of bump characteristics using regionFinder().

When NOT to Use

• For basic linear modeling of independent, unclustered genomic features, use limma because bumphunter is specifically designed to share information between nearby clustered locations.
• For end-to-end analysis of Illumina 450k arrays without manual matrix setup, use minfi because it provides a tailored wrapper around the bumphunter engine.
• For whole-genome bisulfite sequencing (WGBS) data requiring specialized smoothing, use bsseq because it adapts the bump hunting methodology specifically for bisulfite data.

Data Requirements

• Signal Matrix: A numeric matrix (y) of genomic signals where rows represent genomic locations and columns represent biological replicates.
• Design Matrix: A design matrix (X) representing the experimental covariates, created with standard R modeling functions.
• Genomic Coordinates: Vectors for chromosome (chr) and genomic positions (pos) corresponding to the rows of the signal matrix.

Key Parameters

• maxGap (300): Maximum distance (in base pairs) between genomic positions to be grouped into the same cluster in clusterMaker().
• cutoff (0.05 or 0.5): Numeric threshold determining the boundary for "positive" or "negative" segments in getSegments() and bumphunter().
• B (250): Number of permutations used to assess uncertainty and create a null distribution in bumphunter().
• verbose (TRUE): Logical to print progress information during parallel bumphunting.
• cores (2): Number of parallel backend cores registered via registerDoParallel().

Best Practices

• Group genomic locations into distinct units using clusterMaker() before running segment-finding functions.
• Use the doParallel package and registerDoParallel() to distribute permutation computations across multiple cores.
• Ensure the design matrix (X) contains an intercept term and the covariate of interest; avoid using permutation testing if adjusting for multiple confounders.
• Run foreachCleanup() after parallel execution to properly close connections.

Common Pitfalls

• Slow execution during permutation testing; fix this by setting up a parallel backend with registerDoParallel() before calling bumphunter().
• Permutation test warnings when adjusting for confounders; fix this by noting that permutation testing is not recommended when the design matrix has columns other than the intercept and primary covariate.
• Locations on different chromosomes being clustered together; fix this by ensuring the chr vector is correctly passed to clusterMaker(), which strictly separates chromosomes.

Alternatives

• limma: Provides lmFit for linear modeling of biological replicates without the spatial smoothing and clustering steps.
• minfi: Offers a specialized implementation of the bumphunter methodology tailored specifically for Illumina 450k methylation arrays.
• bsseq: Adapts the bump hunting conceptual approach specifically for whole-genome bisulfite sequencing data.
• charm: Provides modifications of the bump hunting methodology for CHARM-like methylation microarrays.

Citations

• Jaffe, A. E., et al. (2012). Bump hunting to identify differentially methylated regions in epigenetic epidemiology studies. International Journal of Epidemiology, 41(1), 200-209.
• Efron, B., & Tibshirani, R. J. (1993). An Introduction to the Bootstrap.

References

• Homepage: https://bioconductor.org/packages/bumphunter
• Vignette: https://bioconductor.org/packages/release/bioc/vignettes/bumphunter/inst/doc/bumphunter.pdf

---

Run this on BioMate

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

▶ Open bumphunter on BioMate →