CHOP: Phased path extraction from population graphs

How to use

Prerequisites

CHOP requires:

- Python 2.7
- networkx 1.9
- bitarray

Input graphs should be directed and acyclic.

Installing

git clone https://github.com/tomokveld/CHOP
cd CHOP
python setup.py install

You should now be able to call CHOP with:

chop

Example usage

Running CHOP

In this example we run CHOP on a variation graph that encodes three haplotypes:

CHOP is called with k = 4 (-k 4), with haplotyping enabled (-H), running serially (-p 1), with an input graph in the GFA format from the examples directory, and outputting the k-paths in Figure_1_k4.fa.

chop -H -k 4 -p 1 -g examples/Figure_1.gfa.gz -o Figure_1_k4.fa

Interpreting the output

CHOP outputs the haplotyped k-paths in the FASTA file, Figure_1_k4.fa, i.e. the null graph, in which each path is represented as a contig:

>1+0,5_3+0,1_4+0,2_9+0,3
ATCCTATTCAC
>1+2,5_2+0,1_4+0,2_5+0,3_7+0,1_8+0,3_9+0,3
CCTCTTTTCTGCCCAC
>5+0,3_6+0,1_8+0,3
TTCCGCC

How do these paths in the null graph relate back to the original graph? Every path comes in a pair describing an interval (> line) and a corresponding sequence. This interval describe the mapping of the path to corresponding nodes and nucleotide offsets. The notation of these intervals follows as node+start_offset,end_offset with _ separating subsequent nodes.

Below is shown how the null graph (paths are shown in the same order as above) relates back to the original graph. The colored lines represent mapped 4-mers sampled from the original graph onto the null graph.

Read alignment

To map reads onto this null graph, you can pick your favorite linear reference aligner (here shown for BWA). First the null graph has to be indexed:

bwa index Figure_1_k4.fa

Once the null graph is indexed read alignment can follow, e.g.:

bwa mem Figure_1_k4.fa reads.fq > out.sam

The output SAM file can now be used for further downstream processing.

Building compatible graphs from VCF files

At https://bitbucket.org/tomokveld/vcf-graph-gfa/ a script with documentation can be found to build graphs (.GFA) from input an input VCF file and FASTA file.

For quick reference these output GFA files are formatted as:

H   VN:Z:1.0
H   BUILD:Z:VCF2GRAPHGFA
H   ORI:Z:sample1;sample2;REFERENCE
S   1   TTGACCGA...CCCACAG    *   ORI:Z:0;1;2 N:I:1700    START:i:0
S   2   T   *   ORI:Z:1;2   N:I:1   START:i:1700
L   1   +   2   +   0M
S   3   C   *   ORI:Z:0 N:I:1
L   1   +   3   +   0M
S   4   CATCGCCA...TGGAAA *   ORI:Z:0;1;2 N:I:119 START:i:1701
L   2   +   4   +   0M
L   3   +   4   +   0M
S   5   G   *   ORI:Z:1;2   N:I:1   START:i:1820
L   4   +   5   +   0M
S   6   T   *   ORI:Z:0 N:I:1
L   4   +   6   +   0M

Surjecting graph alignments SAM files back to a linear reference

To simplify downstream analysis it can be helpful to map back to a single reference... TODO

Projecting alignments back onto the variation graph

In src/simple_chop_coverage.py we give a simple example to map back alignment coverage onto the graph.

Command line parameters

-h, --help Show a help message and exit
-g GFA, --gfa GFA Input GFA file
-k KMER k-mer value
-o OUTPUT Output file to write results
-H, --haplotype Use haplotype information in the graph
-p PROCESSES Number of processes to start (serial is default p = 1)
-s PREFIX, --prefix PREFIX Add an optional prefix for each path entry
-e, --edge Use if the haplotype annotations are on the edges rather than the nodes
-l {DEBUG,INFO,WARNING,ERROR,CRITICAL} Set the logging level, for debugging purposes
--version Show program's version number and exit

License

This project is licensed under the MIT License - see the LICENSE.md file for details.

tomokveld / CHOP