redndgreen8 / hmcnc

Hidden Markov Model based Copy number caller

hmcnc - Hidden Markov Copy Number Caller

Pipeline for calling CNVs in assemblies or alignments

---

---

Initially the required packages need to be installed. On our linux cluster the easiest package management software is Anaconda/Miniconda.

1. Download shell script (64bit):

https://docs.conda.io/en/latest/miniconda.html#linux-installers

2. Run script and setup channels

bash Miniconda3-latest-Linux-x86_64.sh

https://docs.conda.io/projects/conda/en/latest/user-guide/install/linux.html

conda config --add channels defaults

conda config --add channels bioconda

conda config --add channels conda-forge

3. Project env - There are many ways to do this but you can set up a project specific environment with all the packages you need.

https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html#creating-an-environment-with-commands

Required packages

• bedtools
• samtools
• snakemake
• boost
• R
• gxx
• tabix

conda create --name <proj_env> bedtools samtools snakemake boost R tabix

conda install can be used to further add packages to environment with explicit version numbers.

conda install -n <proj_env> scipy=0.15.0

Always activate the env before attempting a run

conda activate <proj_env>

You might run into a conda init error the first time so run conda init and rerun

Compiling cpp source files

You can run the snakemake based make file:

snakemake -s make.smk.py --config boost=<boost> -j 1 -p

where <boost> is the location of boost_install/include folder.

Most likely {anaconda install}/envs/{proj_env}/include.

Running program

./hmcnc

usage: hmcnc <command> [<args>]

Hidden Markov Copy Number Caller command options:

asm: Run a denovo assembly.

aln: Run a reference alignment.

Alignment

./hmcnc aln -h

usage:

./hmcnc aln --bam <input.bam> --index <ref.index> [<args>]

Run HMM caller on alignment. If available, provide repeat mask annotation (--repeatMask, -r) for the reference used to filter >80 percent repeat content calls.

./hmcnc aln

required arguments:

• --bam BAM Bam file of Alignment, bam index file should be in same dir. (default: None)
• --index INDEX index file of reference/assembly coordinates (default: None)

optional arguments:

• --mq MQ Min MapQ for reads (default: 10)
• --outdir OUTDIR Output directory (default: .)
• --repeatMask REPEATMASK Provide reference based repeat bed file. (default: No)
• --coverage COVERAGE Provide genome-wide coverage, if not specified, caller will calculate mean coverage per contig. (default: No)
• --subread SUBREAD [1|0], Needs subreads filtering or not.(PacBio clr reads) (default: 0)
• -t THREADS, --threads THREADS Threads available (default: 1)
• --epsi EPSI epsilon parameter (default: 90)
• --minL MINL min collapse length (default: 15000)
• --scr SCR Scripts DIR (default: /scratch2/rdagnew/hmmnew/snakemake)

Assembly

./hmcnc asm

Same as above but without repeat mask step.

Main Output

• coverage.bins.bed.gz (coverage in 100bp windows)
• copy_number.tsv (copy number profile of whole genome)
• DUPcalls.copy_number.tsv
• DUPcalls.masked_CN.tsv (calls repeat masked)
• DUPcalls.composite.bed (Bookended calls are merged)
• DUPcalls.masked_CN.composite.tsv
• {GENOME}.noclip.pdf (plot of coverage and copy number across WG)
• DELcalls.copy_number.tsv (del calls are naturally recovered)
• CallSummary.tsv