methylCRF

Combining MeDIP-seq and MRE-seq to estimate single CpG methylation genome wide.

---

Overview

No current assay of DNA methylation provides single-CpG resolution, comprehensive genome-wide coverage, and cost feasibility for a typical laboratory. To address this gap, we introduce methylCRF, a novel Conditional Random Field-based algorithm that integrates methylated DNA immunoprecipitation (MeDIP-seq) and methylation-sensitive restriction enzyme (MRE-seq) sequencing data to predict DNA methylation levels at single CpG resolution. Our method is a combined computational and experimental strategy to produce DNA methylomes of all 28 million CpGs in the human genome for a fraction (<10%) of the cost of whole genome bisulfite sequencing methods.

methylCRF was benchmarked for accuracy against Infinium arrays, RRBS, WGBS sequencing and locus specific-bisulfite sequencing performed on the same embryonic stem cell line. methylCRF transformation of MeDIP-seq/MRE-seq was equivalent to a biological replicate of WGBS in quantification, coverage and resolution.

This page describes how to download the methylCRF-suite of programs and run it on a MeDIP-seq/MRE-seq data set.

---

Usage

Name

• methylCRF.pl - Predict genome-wide CpG methylation levels using MeDIP-seq and MRE-seq assays.

Synapsis

• methylCRF.pl [options] <MeDIP_seq bed> <MRE_seq bed> <MRE virtual digest xx_cpg.bed> >mCRF.bed 2>err

Description

Arguments

MeDIP_seq.bed	MeDIP-seq (with with uniq alignemnts)
MRE_seq.bed	MRE-seq (normalized read counts by MRE_norm.pl)
MRE vdigest	CpG's sampled by MRE-seq. format: CpGid[0|1] where CpGid corresponds to cpg.bed.

Options

-gdir	directory with model specific files: crf.list, cut.list, model files: [crfname].mdl [dflt: mdl]
-gdat	directory with genomic specific files: cpg.bed (Epigenetic Roadmap blacklist removed), gdata.tbl, [crfnm]_cpg.bin [dflt: gdat]
-gap	The size of gap between consecutive CpG's for which to start a new region: 750bp was used for model in paper. [dflt: 750]
-eid	Prefix to add to all output files [dflt: eid]
-sfrom/-gto	Optionally only run certain steps (sfrom: start from, gto: go untill): Generate MRE/DIP avg window score files (1.5 DIPMRE count files) Make MRE-sampled CpG file [deprecated] Generate CRF tables (~ 1.5 hr) Prediction Combine predictions into ensemble make dirs to put extra files in [have to give explicitely]

---

Requirements

1. a working Unix-like OS
2. standard GNU tools and perl.
3. samtools
4. pre-aligned MeDIP-seq and MRE-seq libraries in bam format

---

How-to

1. Get methylCRF code by either:
   • git

     git clone https://bitbucket.org/mscse/methylcrf.git

   • or wget

     wget https://bitbucket.org/mscse/methylcrf/get/fd50dc97b171.zip 
     unzip fd50dc97b171.zip
     



2. Run make to compile binaries and set up links

   cd methylCRF
   make
   

*Note: stay in the methylCRF directory for the following steps


3. Retrieve a methylCRF model
   • H1ES

     wget http://methylcrf.wustl.edu/h1es_mdl.tgz
     tar -zxf h1es_mdl.tgz
     



4. Retrieve species-specific data files
   • hg19

     wget http://methylcrf.wustl.edu/hg19/hg19_gdat.tgz
     tar -zxf hg19_gdat.tgz
     

   • Please email to request other species not listed below.
5. Prepare MeDIP-seq and MRE-seq data
   • Generate from sequencing experiments
     • MeDIP
     Generate sorted bed file of unique alignments. Note that the reads should be virtually extended to expected fragment length. The score field should 1.
     • ex: chr1 <begin> <begin+length> . 1 <strand>
     • MRE
     Generate normalized MRE counts
     • Retreive an MRE fragment file for. Note for 5-enzyme MRE replace '3enz' with '5enz'. (email me for other species)

           wget http://methylcrf.wustl.edu/hg19/MRE_frags/MRE_3enz_4_6000.bed
       

     
     
     • Run MRE_norm.pl. It outputs <expr_prefix>_MRE.bed (as well as some diagnostic files)
       

     ( 
      samtools view <bam1> |sam2bed.pl -r -q <minMapq> -c <bam1call> - 
      samtools view <bam2> |sam2bed.pl -r -q <minMapq> -c <bam2call> - 
     ) | MRE_norm.pl - MRE_{3,5}enz_4_6000.bed <expr_prefix> 

     NOTE: the call number is the number of bases that were cut off the read during sequencing (it's optional)
   • Alternately Download preprocessed H1ES data used for the paper:

     wget http://methylcrf.wustl.edu/h1es_dipmre.tgz
     tar -zxf h1es_dipmre.tgz

6. Estimate Methylation by running methylCRF
   • for example, for the H1ES data set used in the paper:

     export PATH="~/src/methylCRF/:$PATH"
     ./methylCRF.pl -eid H1ES -mdir h1es_mdl -gdat hg19_gdat H1ESmrg_DIP.bed H1ESmrg_MRE.bed MRE_3enz_4_6000_cpg.bin >H1ES_mCRF.bed 2>err
     

     NOTE1: use the 3 or 5 enzyme fragment file use for MRE normalization.
     NOTE2: if some script can't find another, check that you have . in your path (ie, in: echo $PATH)

---

MRE Protocol Enzymes

• 3 enzyme: HpaII (CCGG), SsiI (CCGC), Hin6I (GCGC)
• 4 enzyme: above 3, HpyCH4IV (ACGT)
• 5 enzyme: above 4, BstUI (CGCG)

---

Downloads

Experimental Data

• H1ES
  • MeDIP: unique alignment locations bed
  • MRE: normalized read counts bed

Model

• H1ES

Species Data

• Human
• Dog
• Zebrafish (Note: please remove rmsk_RNA from h1es_mdl/crf.list)
• Mouse
• Rat

---

Future

• As more WGBS data becomes evailable, we plan to retrain methylCRF with combinations of cell-types.
• We will continue to explore the integration of more types of data.

---

---

---

methylCRF Google Group

Please post questions here or send questions directly to methylcrf _at_ googlegroups.com

---