cmuts

Fast, multithreaded pair-HMM counting of MaP-seq mutations.

Web Server

The web server runs cmuts on Hugging Face, bypassing the need for a local install. It is suitable for jobs of a few references and a few thousand reads.

Installing

Each release carries static cmuts binaries for Linux (x86_64, aarch64) and macOS (arm64). These bundle htslib and HDF5, so no libraries need to be installed to run them. Alternatively, you can build from source, which may provide marginal speedups.

cmuts align calls separate programs, which must be on the PATH.

• minimap2
• samtools
• vsearch for paired-end input.

On macOS with Homebrew:

brew install minimap2 samtools vsearch

On Debian and Ubuntu:

apt install minimap2 samtools vsearch

cmuts plot requires a Python 3 installation with the h5py, numpy, and plotly packages.

python3 -m pip install h5py numpy plotly

Usage

Compute reactivity rates via the pair HMM on a specific experiment:

cmuts hmm -f references.fasta -o treated.h5 treated.bam

Subtract rates computed from multiple experiments:

cmuts sub -o reactivity.h5 treated.h5 untreated.h5

Normalize reactivity rates across experiments:

cmuts norm -o apo-normalized.h5 -o holo-normalized.h5 apo.h5 holo.h5

Generate an interactive report of the results:

cmuts plot \
    apo-normalized.h5 --label "Apo" \
    holo-normalized.h5 --label "Holo"

Documentation

See the docs for more details on each program, the HDF5 outputs, and special use cases.

Development

See CONTRIBUTING.md for information on development builds and running tests.