No description, website, or topics provided.
Python Shell
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.
methrafo bamScript Jul 1, 2017
model trained_model sklearn 0.18.1 update Jun 21, 2017
test test script Jun 22, 2017
.travis.yml travis config Jun 22, 2017
LICENSE.txt update1 Apr 12, 2017 README update Jun 22, 2017
pre_install dependency automatic installation Jun 22, 2017 update1 Apr 12, 2017

  __  __      _   _     _____        __      
|  \/  |    | | | |   |  __ \      / _|     
| \  / | ___| |_| |__ | |__) |__ _| |_ ___  
| |\/| |/ _ \ __| '_ \|  _  // _` |  _/ _ \ 
| |  | |  __/ |_| | | | | \ \ (_| | || (_) |
|_|  |_|\___|\__|_| |_|_|  \_\__,_|_| \___/ 

Build Status License: MIT


This python package is designed to estimate genome-wide absolute methylation level based on the MeDIP-Seq data.

For R users, please refer to for R wrapped package.


Linux, MacOS


python 2.7.x

It was installed by default for most Linux distributions and MAC.
If not, please refer to for installation instructions.

Python packages dependencies:
-- scikit-learn
-- scipy
-- numpy
-- pyBigWig

The python script will try to install these packages automatically. However, please install them manually if, by any reason, the automatic installation fails.

External packages dependencies: The software requires the following external packages to process bam files. (The following packages must be installed if you want to calculate methylation level based on raw bam files).


We provided a "pre_install" shell script to install those dependencies automatically. However, please install them manually if the automatic installation fails.


Manual dependency installation (skip this section if the automatic dependency installation succeeded):


First, download the software package.
Second, cd to the package directory.
Third, run script to install.

Linux, Mac:

$ sudo python install 

Or install using pip

$sudo pip install methrafo


The following 4 commands were provided by the methrafo package:
methrafo.bamScript,, methrafo.train, methrafo.predict.

  • 1)
    This command is used to download the genome files.
    Files: chromosome sequences(e.g. chr1.fa.gz), chromosome sizes(e.g. hg19.chrom.sizes)

     $ <reference_genome_id> <output_directory>


     $ hg19 hg19

    reference_genome_id represents the ID of the genome (e.g. hg19,mm10, et al.)

  • 2) methrafo.bamScript
    This command is used to convert bam file to bigWig file (RPKM).

     $methrafo.bamScript <bam_file> <genome size file>

    bam_file => bam file
    genome size file => chromosome sizes .e.g. hg19.chrom.sizes.

     chr1	249250621
     chr2	243199373
     chr3	198022430
     chr4	191154276
     chr5	180915260
     chr6	171115067
     chr7	159138663
     chrX	155270560
     chr8	146364022
     chr9	141213431
     chr10	135534747
     chr11	135006516
     chr12	133851895
     chr13	115169878
     chr14	107349540
     chr15	102531392
     chr16	90354753
     chr17	81195210
     chr18	78077248
     chr20	63025520
     chrY	59373566
     chr19	59128983
     chr22	51304566
     chr21	48129895


     $methrafo.bamScript example/example_raw.bam hg19/hg19.chrom.sizes
  • 3) methrafo.train
    This command is used to train the random forest model.

    $methrafo.train <downloaded_reference_genome_folder> <MeDIPS.bigWig> <Bisulfite.bigWig> <output_prefix>   

    <downloaded_reference_genome_folder>: The downloaded genome reference folder (e.g. hg19).
    <MeDIPS.bigWig> : The bigWig file representing the RPKM on each position of the genome.
    <Bisulfite.bigWig>: The bigWig file representing the Bisulfite-Seq methylation level (ground truth).
    Note: if your input is .bam file, please use methrafo.bamScript to convert it to bigWig format first.


    $methrafo.train hg19 example/ example/ trained_model
  • 4) methrafo.predict
    This command is used to predict the methylation level based on MeDIP-Seq data.


     $methrafo.predict hg19 example/ trained_model.pkl example_out

Command description:

  • (1) methRafo accepts the following formats as the input:
    .bam => mapped reads from MeDIP-Seq result.
    .bw => bigWig file from MeDIP-Seq (representing the RPKM value).

    If the input is in .bam format, you need to use methrafo to convert it to bigWig format. Please refer to methrafo.bamScript command above for instruction.

    You might need to use to download corresponding reference genome. But you are also allowed to download the reference genome yourself. The reference genome folder needs to contain the following files: all chromosome sequences in fasta format; chromosome size file containing the length information of each chromosome.

  • (2) We provided a trained model on human (based on a breast luminal cells dataset from roadmap database). We tested it on a few other datasets on human and it shows pretty good performance in terms of running time and correlation with Bisulfite-Seq data. You can use the trained model we provided or you can use the methrafo.train script to build your own model. Please refer to methrafo.train command for complete details.

  • (3) With the provided trained model (or your own trained model), we provided the command methrafo.predict to predict the genome wide methylation level. The output file is a Wiggle (.wig) format file. You can visualize it using IGV or UCSC genome browser. You can also get the methylation level for any given genomic region easily from the generated wig file.


We provided a test example inside the "test" folder. "test" is the bash script to perform the test on the example dataset.

# download reference genome (if the server is not responding, please download them manually) hg19 hg19

#predict based on provided MeDIP-Seq bam 
methrafo.bamScript example_medip.bam hg19/hg19.chrom.sizes
methrafo.predict hg19 ../model/rr.pkl e1_out

# predict based on provided MeDIP-Seq bigwig 
methrafo.predict hg19 ../model/rr.pkl e2_out


This software was developed by ZIV-system biology group @ Carnegie Mellon University
Implemented by Jun Ding


This software is under MIT license.
see the LICENSE.txt file for details.


zivbj at
jund at