methylFlow

Cell-specific methylation pattern reconstruction. Currently uses an LP formulation and solver. The lemon graph library http://lemon.cs.elte.hu/trac/lemon and the glpk LP solver http://www.gnu.org/software/glpk/ are included in this repository. The ezOptionParser.hpp is also included http://ezoptionparser.sourceforge.net/

The algorithm is described and tested in this publication: http://bioinformatics.oxfordjournals.org/content/32/11/1618.abstract

Installation

This project uses cmake for building and requires at least version 2.6. It also uses c++11 so use a compiler that supports this (e.g., g++ >= 4.7 or clang >= 3.4)

$ git clone https://github.com/hcorrada/methylFlow.git
$ cd methylFlow
$ git submodule init
$ git submodule update
$ mkdir build && cd build
$ cmake ..
$ make
$ make install

To compile with DEBUG flags use

...
$ mkdir build_devel && cd build_devel
$ cmake -DCMAKE_BUILD_TYPE=Debug ..
$ make
...

Usage

MethylFlow: methylation pattern reconstruction

USAGE: methylFlow -sam -i reads.sam -o mfoutput [OPTIONS]

OPTIONS:

-chr, -Chr ARG                    chr name for tsv files, not required for sam
                                  input file

-cpgloss, -p, -P, --cpgloss       Use cpg-loss instead of region-loss.

-e, -eps, -E, --eps ARG           Regularization parameter search threshold.

-end, -End, --end ARG             Only process reads aligning before given
                                  location.

-h, -help, --help, --usage        Display usage instructions.

-i, -in, --in, --input ARG        Read input file. Default:Tab-separated format:
                                  start length strand
                                  methyl(offset[M|U]
                                  substitutions(ignored))

-l, -lam, -lambda, --lambda ARG   Regularization parameter value.

-o, -out, --out, --output ARG     Output directory. Directory must exist before
                                  running. Files written: cpgs.tsv, components.tsv,
                                  patterns.tsv, regions.tsv

-s, -scale, -S, --scale ARG       Scale parameter value.

-sam, -SAM, --sam                 Input file is in SAM format instead of default
                                  tab-separated format.

-start, -Start, --start ARG       Only process reads aligning after given
                                  location.

-v, -verbose, -V, --verbose       Verbose option.

EXAMPLES:

methylFlow -sam -i reads.sam -o mfoutput -l 10.0 -s 30.0 -e 0.1

Output

Upon running, the output directory (mfoutput in the example above) will contain three files with the following format:

cpgs.txt

Tab-separated file of coverage and methylation calls per cpg. Columns

• chr: chromosome name
• pos: cpg position
• Cov: number of reads overlapping CpG
• Meth: number of reads indicating CpG is methylated

components.tsv

Tab-separated file of components found by algorithm. A component is a connected region graph based on overlapping reads. Genomic regions are covered by a single component, thus, cell-specific patterns estimated in a given genomic region are obtained from (one or morei non-overlapping) components that overlap that region.

Columns:

• chr: chromosome name of genomic region covered by connected component
• start: starting position of genomic region covered by connected component
• end: ending position of genomic region covered by connected component
• cid: component id, identifier given to component, used to connect to regions and patterns in other output files
• npatterns: number of cell-specific methylation patterns estimated from this connected component.
• total_coverage: total number of reads overlapping this component's genomic region
• total_flow: the sum of all estimated abundances (flows) for patterns in this region

patterns.tsv

Tab-separated file of cell-specific methylation patterns estimated by methylFlow.

Columns:

• chr: chromosome name
• start: start position of pattern
• end: end position of pattern
• cid: component id, corresponds to id of a component in file components.tsv
• pid: pattern id, identifier given to pattern (unique across patterns within the same component)
• abundance: abundance estimated for this pattern
• methylpat: comma-separated list of methylation status entries of cpgs within pattern. Entries are pos:[M|U] where position is the location of the CpG from the start of the pattern and M|U indicates if the CpG is methylated or unmethylated respectively
• regions: comma-separated list of regions included in pattern (see file regions.tsv)

regions.tsv

Tab-separated file of regions that make up the region graph used in the estimation algorithm. Reads are assigned to a region if they have no disagreement on their methylation pattern. That is, regions contain the longest stretches of overlapping reads with unambiguous methylation patterns.

Columns:

• chr: chromosome name
• start: start position of region
• end: end position of region
• cid: component id, corresponds to identified of component in file components.tsv
• rid: region id, identifier given to region (unique across regions within the same component)
• raw_coverage: number of reads assigned to the region
• norm_coverage: normalized region coverage
• exp_coverage: the sum of abundances of all patterns that include this region
• methylpat: methylation pattern of region, given in same format as patterns.tsv

##Authors

Hector Corrada Bravo hcorrada@gmail.com
Faezeh Dorri

Center for Bioinformatics and Computational Biology
University of Maryland
http://www.cbcb.umd.edu/~hcorrada