ChromHMM Chromatin State Inference

Overview

This skill enables comprehensive chromatin state analysis using chromHMM for histone modification ChIP-seq data. ChromHMM uses a multivariate Hidden Markov Model to segment the genome into discrete chromatin states based on combinatorial patterns of histone modifications.

Main steps include:

Refer to Inputs & Outputs to verify necessary files.
Always prompt user if required files are missing.
Always prompt user for genome assembly used.
Always prompt user for the bin size for generating binarized files.
Always prompt user for the bin size for the number of states the ChromHMM target.
Run chromHMM workflow: Binarization → Learning.

When to use this skill

Use this skill when you need to infer chromatin states from histone modification ChIP-seq data using chromHMM.

Inputs & Outputs

Inputs

(1) Option 1: BED files of aligned reads

<mark1>.bed
<mark2>.bed
... # Other marks

(1) Option 2: BAM files of aligned reads

<mark1>.bam
<mark2>.bam
... # Other marks

Outputs

chromhmm_output/
  binarized/
    *.txt 
  model/
    *.txt
    ... # other files output by the ChromHMM

Decision Tree

Step 0: Initialize Project

Call:

mcp__project-init-tools__project_init

with:

sample: all
task: chromhmm

Step 1: Prepare the `cellmarkfile` (skip this step if signal files are provided)

Prepare a .txt file (without header) containing following three columns:
- sample name
- marker name
- name of the BED/BAM file
- control file of the sample (only provided if the input/control file is available)
example of the cellmark.txt file

cell1    mark1    cell1_mark2.bam    cell1_control.bam
cell1   mark2    cell1_mark2.bam    cell1/control.bam

Step 2: Data Binarization

For BAM inputs:
Call:
- mcp__chromhmm-tools__binarize_bam with:
- path_chrom_sized: Provide by user or detect from the working directory
- input_dir: Directory containing BAM files
- cellmarkfile: Cell mark file defining histone modifications
- output_dir: (e.g. binarized/)
- bin_size: Provided by user
For BED inputs:
Call mcp__chromhmm-tools__binarize_bed instead.
For Signal inputs:
Call: mcp__chromhmm-tools__binarize_signal with:
- input_dir: Directory of signals
- output_dir: (e.g. binarized/)

Step 3: Model Learning

Call

mcp__chromhmm-tools__learn_model