Awesome

covviz

Coverage visualization; a many-sample coverage browser.

The aim of covviz is to highlight regions of significant (passing the user's z-score threshold) and sustained (beyond user specified distance) deviation of coverage depth from the majority of samples. Significance is determined using z-scores for all samples at all points using median absolute deviation. In order for regions to be highlighted, points must be significant consecutively throughout a user specified distance.

If you are analyzing a low number of samples, deviation may be irrelevant. In this case, we can set --min-samples to be greater than our sample total to skip Z-threshold calculation and plot coverages for all samples at all points.

Getting started

From alignments (.bam and/or .cram)

Alignments must be indexed. The input for the covviz workflow are the indexes of the alignments. For BAM, that would be .bai, and .crai for CRAM. Indexes can be generated using samtools on your sorted alignments:

samtools index mybam.bam
# generates mybam.bam.bai

Installation and usage

Install Nextflow if you don't already have it. The only dependency is Java 8 or later, then you run:

curl -s https://get.nextflow.io | bash

Or via Bioconda using:

conda install -c bioconda nextflow

Full nextflow installation instructions are available at: https://www.nextflow.io/

There is no need to download the covviz code prior to execution or any software dependencies when using a container service like Docker or Singularity.

Docker/Singularity

To simplify prerequisite software installations and software version tracking, we strongly recommend running covviz using Docker or Singularity. Docker installation instructions for your operating system are available at: https://docs.docker.com/install/

Then, with Docker or Singularity we run:

nextflow run brwnj/covviz -latest -profile docker \
    --indexes 'data/indexes/*.crai' \
    --fai data/g1k_v37_decoy.fa.fai \
    --gff data/Homo_sapiens.GRCh37.82.gff3.gz

Which gives us ./results/covviz_report.html.

Required arguments

--indexes
- quoted file path with wildcard ('*.crai') to cram or bam indexes
--fai
- file path to .fai reference index

A complete list of arguments can be displayed using:

nextflow run brwnj/covviz -latest --help

Nextflow arguments

In the example above -latest pulls whatever the latest covviz code exists on GitHub prior to execution and -profile docker sets -with-docker within Nextflow.

Other notable options are -resume, which when running a workflow a second will start where previous runs of the workflow left off; and -work-dir which sets the location of all intermediate files generated throughout the workflow.

From coverage intervals (.bed)

The covviz CLI accepts bed3+ as input. If you've already generated your coverage files you can start here and not the Nextflow workflow.

If you would prefer to run indexcov yourself across your .bai or .crai files, the workflow above simply runs:

fai=data/g1k_v37_decoy.fa.fai
goleft indexcov --directory myproject --fai $fai *.crai

This will generate the expected inputs in their anticipated formats for the covviz CLI.

Expected file format

To analyze your coverage data it needs to be in bed3+ format and include a header with sample IDs. The first three column headers are agnostic, but for samples test_sample1, test_sample2, and test_sample3, this would look like:

#chrom   start   end   sample1   sample2   sample3

Installation of CLI and usage

To install the covviz Python package use:

pip install -U covviz

Then CLI usage is:

covviz $bed

A complete list of arguments can be displayed using:

covviz --help

Adding custom metadata (.ped)

There is support for non-indexcov .ped files, though you may have to change the default column IDs pertaining to the column which contains the sample ID and the sex of the sample.

covviz --ped $ped --sample-col sample_col --sex sex_col $bed

Adding annotation tracks

significant_regions

Currently we support GFF, VCF, and BED. GFF tracks are added using --gff where features are 'gene' and attributes have 'Name='. Feature type and attribute regex can be configured using --gff-feature and --gff-attr.

VCF tracks (v4.1) are added with --vcf with the entire INFO string being displayed by default. Specifying --vcf-info with something like 'CLNDN=' will grab just that field when using ClinVar variants. Including large INFO strings for all variants can dramatically increase the size of the covviz report.

Region based annotation tracks can be added using --bed. The name field will be used to identify the regions when present.

Annotation tracks, --gff, --vcf, and --bed, may be specified multiple times.

In all cases, 'chr' will be stripped from the chromosome names.

Interpreting the output

Interactive example

See: https://brwnj.github.io/covviz/

Scaled chromosome coverage

Significant regions will be displayed in color atop a gray region which represents the upper and lower bounds of a given point minus any values deemed significant.

significant_regions

When plotting fewer samples than --min-samples, the gray area plot will not be displayed. Instead, all sample plot traces will be shown.

min_samples

Proportions covered

proportional_coverage

The metadata table will be displayed below the plots.

Interaction

Clicking on plot traces highlights the line and searches the metadata. Double-clicking de-selects lines, resets the plot, and de-selects samples from the table. Clicking on the gene track launches a search for the gene's respective Gene Card. In cases where genes overlap, multiple windows/tabs will be opened.