Before you use the EDACC BED alignment file for anything including the
fragment length estimation and MACS peak/signal calls you need to make
some changes to the BED files. 

Make sure you (1) trim reads to 36 bp and (2) filter reads uniformly
using a 36 bp mappability track since the EDACC BED files artificially
extend reads to 200 bp.

(1) Here is my code for trimming and correcting EDACC files (they dont
    use 0 based starts as required by BED format)
http://www.broadinstitute.org/~anshul/projects/roadmap/src/autocorrect_adjustReadLen.sh
Use it as follows
./autocorrect_adjustReadLen.sh [EDACCBed.gz file] 36 | gzip -c >
    newFile.tagAlign.gz

(2) Also here is the read filtering code.
You will need the following files from this directory
http://www.broadinstitute.org/~anshul/projects/roadmap/src/

filterUniqueReads # This is the main utility that does the filtering

filterUniqueReads needs the MATLAB run-time library installed (this is
free .. see below) and requires mappability and sequence tracks for
filtering.
The MATLAB run time library installation instructions are here
https://code.google.com/p/align2rawsignal/#1._MCR_Installation 
Also after you install you need to set appropriate paths as explained
here https://code.google.com/p/align2rawsignal/#2._Setting_paths

The mappability tracks are here
http://www.broadinstitute.org/~anshul/projects/encode/rawdata/umap/encodeHg19Male/globalmap_k20tok54/
# Copy over this directory exactly as is and maintain the directory
name i.e. globalmap_k20tok54

The sequence tracks are here
http://www.broadinstitute.org/~anshul/projects/encode/rawdata/sequence/encodeHg19Male/
# Copy over this directory exactly as is

mapFilterTagAlignFiles.sh # wrapper shell script around the
filterUniqueReads code. See how filterUniqueReads is called in this
script

slopBed is part of the BedTools suite
https://github.com/arq5x/bedtools2 . The script also uses
bedGraphToBigWig which is part of the UCSC source tools.

Also there is a step where we merge multiple replicates/datasets as
required and subsample reads to a fixed depth. This is standard unix
commands such as zcat and shuf.

(3) You then use the cross correlation code to estimate fragment
    lengths and QC scores. Use the code here
    https://code.google.com/p/phantompeakqualtools/ .
 
(4) The script for MACS2 narrow peak calling and signal generation is
    here
    http://www.broadinstitute.org/~anshul/softwareRepo/macs2.signal.lsf.submitscript.sh 
You should be able to find the specific macs2 commands with the
    parameters in that file. The strand shift is 1/2 the estimated
    fragment length.

(5) My ChromHMM scripts are here
    /broad/compbio/anshul/projects/encode/preprocessing/segmentations/chromhmm/scripts
chmm_binarize.sh: will binarize the tagAlign files
chmm_predict.sh: will infer states based on an existing model
chmm_browser.sh: will create browser tracks
denseBedtoBigBed.sh: will create bigBed tracks from the dense.bed
    state tracks.