DMseg 1.0.2

Python package for DMseg: detecting differential methylation regions in DNA methylome data

DMseg: detecting differential methylation regions (DMRs) in methylome

This is a Python package to search through methylome-side CpGs sites for DMRs between two biological conditions. The algorithm executes the following analysis steps:

1. A linear regression model is fitted to the beta values for each CpG, using the user-input covariates: the first variable is the group label of interest.
2. Nominal p-values from individual CpG associations are used to define the DMR: more than or equal to two consecutive CpGs with p-value <0.05 will be considered as candidate DMR. A likelihood ratio statistic (LRT) is computed for a candidate DMR.
3. Group labels will be permuted for B times, step 1 and 2 are repeated for each permuation dataset. Family-wise error rate is computed using the null distribution of LRT based on permutation.

To install the package:

python -m pip install DMseg

To run the python package, one can first port in the package and the core function file DMseg.py, then call the pipeline function DMseg_pipeline

import DMseg
from DMseg import DMseg
result = DMseg.DMseg_pipeline(betafile, colDatafile, postionfile, maxgap=500, sd_cutoff=0.025, beta_diff_cutoff=0.05, zscore_cutoff=1.96, B=500, seed=1001)

The inputs for the function DMseg_pipeline are

• betafile: the file location for the n x p matrix of methylation beta values (n is the number of samples, p is the number of CpG sites)
• colDatafile: the file location for the covariates for regression analysis
• postionfile: the file location for the chromosomal positions of CpGs
• maxgap is the maximal base pairs between two adjacent CpGs that can be considered as within the same DMR
• sd_cutoff is the minimal inter-sample standard deviation for a CpG to be considered for differential methylation
• beta_diff_cutoff is the minimal mean differences between the two comparison groups for a CpG to be considered for differential methylation
• zscore_cutoff is the minimal z-score for a CpG to be considered for differential methylation
• B is the number permutations to compute family-wise error rate
• seed is the seed for the random number generator when permuting the group labels