gCNV Analysis Tools

Some analysis tools for the exome GATK-gCNV pipeline. This is NOT the GATK-gCNV pipeline itself. Most of these workflows rely on post-processing steps internal to the Talkowski lab and are not applicable to the output of GATK GermlineCNVCaller. They are also designed to be run on Terra.

There is a WDL workflow for each analysis and an R script that contains the core analysis code. It is possible to run the analysis just using the R script if the inputs are formatted correctly.

Repository Structure

Path	Description
src/scripts/	core analysis scripts
src/gelpers/	R package with helper code
wdl/	WDL workflows
Dockerfile	Dockerfile for all the workflows
Makefile	Makefile with some useful recipes

Workflows

There are workflows for the main analysis and some helper workflows to help construct some of the inputs for the main workflows.

MakeSampleBatchMap

• workflow: wdl/MakeSampleBatchMap.wdl

The purpose of this workflow is to create a file that links batch IDs to sample IDs. The batch ID refers to the ID that is in the final GATK-gCNV callset, identifying the samples that were processed together using a single model. Typically it is derived from the sample set IDs on Terra. The workflow should be run once per sample set. The outputs are a TSV with batch IDs in the first column (the same ID repeated) and sample IDs in the second column, and the batch ID as a string.

ConcatSampleBatchMaps

• workflow: wdl/ConcatSampleBatchMaps.wdl

This workflow should be run on a sample set set and simply concatenates all the files produced by MakeSampleBatchMap for each of the constituent sample sets.

FixDCR

• workflow: wdl/FixDCR.wdl

Modifies an old version of the batch-level denoised copy ratio (dCR) file to be compatible with these workflows. Probably will not be needed.

MergeDcr

• workflow: wdl/MergeDcr.wdl

Merges per-sample dCR files into per-batch dCR files. Probably will be not be needed.

Common Inputs

There are some inputs that are commonly used and are described here.

callset

The callset refers to the final table of CNV calls produced by the GATK-gCNV pipeline.

pedigree

Pedigree file based off the GATK PED format. The GATK PED format is described here. Briefly, it is a six column TSV describing the family ID, sample ID, paternal ID, maternal ID, sex, and phenotype, in that order. The sex is 1 for male and 2 for female. All other values are treated as unknown. The phenotype is 1 for unaffected and 2 for affected. All other values are treated as unknown.

intervals

This is the file of genomic intervals used to collect read counts. It is a three column TSV with the chromosome, start, and end of each intervals. All intervals are 1-based start and inclusive end. The intervals must be disjoint.

dcr_files and dcr_indices

The dCR and dCR index files. These are usually the per-batch ones.

batch_ids

The IDs used to identify batches of samples that used the same CNV calling model. These must match the batch IDs in callset.

sample_set_ids

The IDs used for the sample sets on Terra.

sample_batch_map

A two-column TSV of the format produced by ConcatSampleBatchMaps.

runtime_docker

The URI for the Docker image created from Dockerfile. All of the workflows use the same Docker image.

de novo Calling

Annotate de novo CNVs in a callset.

• workflow: wdl/AnnotateDeNovo.wdl
• script: src/scripts/annotate_denovo_cnv.R

Inputs

recal_freq

Should the variant site frequencies be recalibrated? The default is to set the site frequency of every variant to the frequency just among the parents. If false, there must be a column named "sf" in the callset with the site frequencies.

hq_cols

A comma-separated list of column names in the input callset that must all be true (as parsed by data.table::fread) for a CNV to be considered high quality. This is relevant as only high quality variants will be checked for de novo status. The default is PASS_SAMPLE,PASS_QS.

max_freq

The maximum site frequency (as a decimal fraction) a CNV can have. All variants above this threshold will be ignored. The default is 0.01.

skip_allosomes

Should CNVs on allosomes be skipped. The default is to not skip. It is not necessary to set this option just because there aren't any CNVs on allosomes in the callset.

Description

The source code is the best place to look for the de novo calling steps. Briefly, there is a check for overlap between CNVs in the parents and the child and then further filtering of those without overlap using the dCR values.

Outputs

The input callset file will have a column added with the predicted inheritance status of each variant: denovo, inherited, mosaic_father, etc.

Autosomal Ploidy Calling

Predict the ploidy of autosomes.

• workflow: wdl/CheckPloidy.wdl
• script (ploidy calling): src/scripts/check_ploidy.R
• script (merging ploidy tables): src/scripts/merge_ploidy_matrices.R

Inputs

contigs

The list of chromosomes to check for aneuploidies.

is_hg19

Are the chromosomes in the dCR files named according to the hg19 naming scheme (i.e. 1, 2, etc).

Description

For each chromosome in contigs:

1. dCR values are extracted from the dCR files
2. joined across all the batches by interval coordinates
3. grouped into 36 bins
4. the mean sample dCR is computed for each bin and plotted across the chromosome
5. the mean sample dCR across all the bins is computed and used as the predicted ploidy
6. any sample with predicted ploidy less than 1.5 or greater than 2.5 is treated as having an aneuploidy
7. any sample with a predicted aneuploidy has its dCR intervals (the original from the dCR files) grouped into 101 bins and those are plotted
8. the table of the predicted ploidy at the chromosome, the plots of ploidy and any predicted aneuploidies are collected in the output

The workflow combines the outputs across contigs.

Outputs

ploidy_plots

A tar file with the ploidy plots, including any aneuploidies.

ploidy_matrix

A TSV with the predicted ploidy of every sample/chromosome pair.

aneuploidies

A TSV with the predicted ploidy of any sample/chromosome pair predicted to be an aneuploidy.

Allosomal Ploidy Calling

Predict the ploidy of allosomes. Requires that the intervals used to run GATK-gCNV contained intervals on the allosomes.

• workflow: wdl/CheckSex.wdl
• script: src/scripts/check_sex.R

Inputs

Nothing unique to this workflow.

Description

Predicts the ploidy of the allosomes by computing their mean dCR.

Outputs

A TSV of the predicted ploidy of every sample at chrX and chrY.

CNV Evidence Plotting

Make plots of CNV evidence.

• workflow: wdl/PlotCNVEvidence.wdl
• script: src/scripts/plot_cnv_evidence.R

Inputs

calls_per_split

Number of CNVs to process per shard.

Description

Plots a sample's dCR values for every interval overlapping a CNV.

Outputs

A tar file of plots.

de novo CNV Evidence Plotting

Make visualizations of potential de novo CNVs.

• workflow: wdl/PlotDeNovoEvidence.wdl
• script: src/scripts/plot_denovo_evidence.R

Inputs

denovos

GATK-gCNV callset with inheritance status annotated.

Description

Plots the dCR values for every interval overlapping a CNV for all members of a trio.

Outputs

A tar file of plots.