Functional Segmentation of the Methylome

Implements FUSE (Functional Segmentation of DNA methylation data), a data-driven method for identifying spatially coherent DNA methylation segments from whole-genome bisulfite sequencing (WGBS) count data. The method performs hierarchical clustering of CpG sites based on methylated and unmethylated read counts across multiple samples and determines the optimal number of segments using an information criterion (AIC or BIC). Resulting segments represent regions with homogeneous methylation profiles across the input cohort while allowing sample-specific methylation levels. The package provides functions for clustering, model selection, tree cutting, segment-level summarization, and visualization. Input can be supplied as count matrices or extracted directly from 'BSseq' and 'methrix' objects.

methFuse

methFuse implements FUSE: FUnctional SEgmentation of DNA methylation data through hierarchical clustering.

Features

• Hierarchical clustering based on methylation count matrices
• Detection of optimal number of clusters
• Summarization of segments, including stability flag
• Per-segment methylation estimates per sample

Installation

From GitHub (Development Version)

Either using remotes: (recommended)

# Install remotes if needed
install.packages("remotes")

# Install fuseR from GitHub
remotes::install_github("holmsusa/methFuse")

or using devtools:

# Install devtools if needed
install.packages("devtools")

# Install fuseR from GitHub
devtools::install_github("holmsusa/methFuse")

System Requirements

• R version ≥ 4.0
• C++ toolchain for native code compilation

You may need platform-specific tools:

• macOS: Xcode Command Line Tools (xcode-select --install)
• Linux: build-essential
• Windows: Rtools

Supported input types

fuse.segment() supports the following input formats:

• Matrix input

  • Unmethylated counts (K0)
  • Methylated counts (K1)
  • Explicit chromosome (chr) and position (pos) vectors

• BSseq objects (Bioconductor)

  • Counts and genomic coordinates are extracted automatically
  • Requires bsseq package

• methrix objects (Bioconductor)

  • Supports large datasets via DelayedMatrix
  • Counts and genomic coordinates are extracted automatically
  • Requires methrix and DelayedArray packages

Install needed packages with

BiocManager::install(c("bsseq", "methrix", "DelayedArray"))

Quick Example

library(fuseR)
set.seed(1234)

# Generate sample data
# Unmethylated counts, T's
K0 <- matrix(
  rep(c(sample(0:20, 200, replace = TRUE), sample(20:40, 200, replace = TRUE)), 2),
  nrow = 100, byrow = TRUE
)
# Methylated counts, C's
K1 <- matrix(
  rep(c(sample(20:40, 200, replace = TRUE), sample(0:20, 200, replace = TRUE)), 2),
  nrow = 100, byrow = TRUE
)

# Perform segmentation
segment_result <- fuse.segment(
  K0, K1, 
  chr = sub("\\..*$", "", rownames(K0)), 
  pos = as.numeric(sub("^.*\\.", "", rownames(K0)))
)

# Access summary and per-segment betas
head(segment_result$summary)
head(segment_result$betas_per_segment)

Check out a full example workflow in the vignette.

License

This package is licensed under the MIT License. See LICENSE for details.

Author

Susanna Holmstrom