ClearCNV: Clinical sequencing panel CNV caller and visualizer

• Code Formatting: black

Installation

conda

ClearCNV is available on conda: https://anaconda.org/bioconda/clearcnv

I'd recommend to create a conda env:

mamba create -n clearcnv clearcnv -c conda-forge -c bioconda

or

conda create -n clearcnv clearcnv -c conda-forge -c bioconda

Then clone this repo to your favorite location git clone git@github.com:bihealth/clear-cnv.git and cd clear-cnv into it. Lastly, activate the environment via mamba activate clearcnv or conda activate clearcnv. Now you can run the commands listed below.

Quickstart

Case one

You have one gene panel (e.g. called '/path/to/genepanel.bed') and a collection of aligned short reads (sample_xy0.bam, sample_xy1.bam, ..) and you want to call CNVs.

• Write a 'meta-file' like this one.
  • Copy all full paths of your bam files to a txt-file e.g. '/path/to/bams.txt'.
  • Your '/path/to/meta.tsv' file would look like this: genepanel\t/path/to/bams.txt\t/path/to/genepanel.bed
• Use clearCNV workflow_cnv_calling. Type clearCNV workflow_cnv_calling --help to see how.
• Check the QC files to see if everything went well.
• Read more: https://github.com/bihealth/clear-cnv/blob/master/README.md#how-to-and-workflow

Case two

You have several panels and you're not really sure if the bam files are assigned correctly to each panel. You want the panels and batches separated and to call CNVs on each of them.

• Write a 'meta-file' like this one.
  • Copy all full paths of your bam files that you think belong to panel 1 to a txt-file e.g. '/path/to/p1_bamfiles.txt'.
  • Copy all full paths of your bam files that you think belong to panel 2 to a txt-file e.g. '/path/to/p2_bamfiles.txt'. Do that for all panels.
  • Your '/path/to/meta.tsv' file would look like this: genepanel\t/path/to/bams.txt\t/path/to/genepanel.bed
• Run clearCNV workflow_reassignment. Type clearCNV workflow_reassignment --help to see how.
• Run clearCNV visualize_reassignment. Type clearCNV visualize_reassignment --help to see how. You'll need to open the URL with your browser.
• After you ran each step in your browser, there will be a folder that contains all newly assigned batches. In each panel/batch you'll find a txt file that contains patchs to .bam files. These are your batches! Proceed with the clearCNV workflow_cnv_calling step for each batch. Type clearCNV workflow_cnv_calling --help to see how.
• Read more: https://github.com/bihealth/clear-cnv/blob/master/README.md#how-to-and-workflow

Quick run checks and examples

Sample reassignment:

Create all files

Execute the shell commamd (from within the cloned repo directory): clearCNV workflow_reassignment --workdir tests/testdata/ --reference tests/testdata/test_reassignment_ref.fa --metafile tests/testdata/test_reassign_meta.tsv --coverages tests/testdata/test_reassignment_coverages.tsv --bedfile tests/testdata/test_reassignment_union.bed --cores 2

• INPUT: working directory given by --workdir, the files given by --reference and --metafile.
• OUTPUT: files created at --coverages and --bedfile. They are used in the next step.

If you want to create the necessary files for yourown data just edit the meta.tsv file analogously to the example at clearCNV/tests/testdata/meta.tsv, where you can add more rows for each targets file (BED-file). It is recommended to use absolute paths in the meta file.

Optionally, drmaa can be used, if the two flags are present: --drmaa_mem 1600 --drmaa_time 4:00, where drmaa is given 16 Gb memory per core and and four hours maximum running time. Also, a cluster config file in .json format can be given with --cluster_configfile config.json

Visualize sample reassignment:

Visualize and adjust the clusterings and final panel assignments

Execute the shell commamd (from within the cloned repo directory): clearCNV visualize_reassignment --metafile tests/testdata/meta.tsv --coverages tests/testdata/cov_reassignment.tsv --bedfile tests/testdata/reassignment_union.bed --new_panel_assignments_directory tests/testdata/panel_assignments

• INPUT: files given by --metafile, --coverages and --bedfile.
• OUTPUT: files found in given directory --new_panel_assignments_directory.

CNV calling

Match scores

At first, match scores are claculated. Go to the directory clear-cnv/ and execute the shell command:

clearCNV matchscores -p testpanel -c tests/testdata/cov.tsv -m tests/testdata/matchscores.tsv

This creates a match score matrix which is used in the CNV calling step.

CNV calls

Now execute this shell command:

clearCNV cnv_calling -p testpanel -c tests/testdata/cov.tsv -a tests/testdata/testpanel/analysis -m tests/testdata/matchscores.tsv -C tests/testdata/testpanel/results/cnv_calls.tsv -r tests/testdata/testpanel/results/rscores.tsv -z tests/testdata/testpanel/results/zscores.tsv -g 15 -u 3

This creates the file tests/testdata/testpanel/results/cnv_calls.tsv which shows one called deletion. if you copy & paste this for your own data, please don't use the -g 15 -u 3 configuration. We use these in here just to be able to work with a tiny example.

More files for analysis can now be found in tests/testdata/testpanel/analysis.

HOW TO and WORKFLOW

clearCNV comprises of two major workflows and three major commads:

workflow

1. re-assignment (not necessary for CNV calling)

   a) clearCNV workflow_reassignment

   b) clearCNV visualize_reassignment

2. CNV calling

   a) clearCNV workflow_cnv_calling

preparations

Some files have to be acquired or created before these commands can be run:

1. re-assignment:

   a) For each sequencing panel a .bed file is needed following this form.

   b) For each sequencing panel (or .bed-file containing all target informations) a simple list of the according .bam files is needed. An example can be found here. Make sure to use absolute paths for this file.

   c) meta-file. This file is a tab-separated file and one example can be found here. To avoid any confusion, we recommend using absolute paths here again.

2. CNV calling:

   a) A genome reference file. It must be the same that was used to create the read alignment files (.bam files).

   b) workflow_cnv_calling does CNV calling for each batch (or sequencing panel associated data set) separately. A text file with all .bam file paths for each batch and panel must be created. Here is an example showing only one .bam file path. Multiple paths are separated with a newline. This file is usually an output of clearCNV visualize_reassignment.

   c) The .bed-file for the sequencing panel for which this batch is put to CNV calling. An example can be found here. Note that gene is optimally replaced with the real name of the exon, gene or target.

   d) A k-mer alignability file in .bed format. Such files can be downloaded from UCSC (e.g. for Hg19 here). A k-mer mappability track can also be created for example using GenMap. In both cases the resulting Wig or BigWig files need to be converted to .bed to be used by clearCNV.

NOTES

The chromosome name scheme in the reference and .bed-file should be of the forms: ChrX, chrX, X or Chr1, chr1, 1.

CNV calling on chr X or chr Y: clearCNV automatically determines the copy number of the gonosomes. If your panel targets only a single gene per chromosome, then it is better to delete according targets from the original .bed file to exclude them. It is necessary to have about double as many samples in your data set to enable meaningful CNV calling on the X or Y chromosomes with roughly equally many women and men in the samples.

If you do sample re-assignment on your own data, followed by CNV-calling, then only one metafile, one coverages file, and one bedfile will be used. This means that --metafile, --coverages and --bedfile are given the same file paths in both workflow steps clearCNV workflow_reassignment and clearCNV visualize_reassignment of clearCNV. The coverages file can not be re-used for the CNV calling steps.

Running Checks

Checks are automatically run on the master branch and pull requests. Unit and integration tests are based on pytest and formatting is enforced with black.

$ make test