Up: Component summary

PanelDoc

Calls CNV ratios/gains/losses using the existing programs PanelDoc (developed by Nord et al. 2011) or CNVPanelizer (Bioconductor package), or both, with custom filtering available. The original scripts have been modified, however no functional changes have been made. Both programs use depth of coverage to call CNVs from targeted sequencing data. It is recommended to study the original documentations prior to running this component. All comments and recommendations described here are based on personal experience from using the programs and does not represent the authors/creators opinions.

All parameters needed for PanelDoc is marked with 'PanelDoc' and all parameters for CNVPanelizer with 'CNVPanelizer'.

PanelDoc

The original description of the scripts implemented in this component is created by Nord et al. and are described in the article 'Accurate and exact CNV identification from targeted high-throughput sequence data, Nord et al., 2011 (Pubmed ID: 21486468)' and github repository: https://github.com/ammawla/PanelDoC. The component also includes some custom made analysis of the PanelDoc output: run_Filtering, run_per_gene and run_Combine_results. To make the usage of the scripts easier, the input structure described in the github readme is different in this anduril component, however the names match the original inputs so the contents are the same.

To run the component you will need a set of input files and two folders. These two folders need to be named 'coverage' and 'Genome'. In the coverage folder you should have all coverage files needed in the anasysis, meaning one coverage file for each region in each sample in the format [Samplename]_[regionname].depth. The extension needs to be .depth, the sample name need to match the sample name in the sample-file (see parameter 'samples') and the region name needs to match the RegionName in the partition file (see parameter 'partition_regions'). For example: Patient01_tissue01_cancer_geneXX.depth. The contents of the file should be three columns: chromosome (format: chr1), position and coverage. Each targeted base needs to be included. In the 'Genome' folder you need to have fasta files for each chromosome separately (chr1.fa, chr2.fa, etc.). The rest of the inputs are defined in the parameters section, but shortly you will need a bedfile, partition file ('partition_regions', this can be created from the bedfile), a list of samples ('samples') and known CNVs ('knowncnvs') and some additional annotation files that can be downloaded from UCSC genome browser ('chain_self', 'refFlat'). You will also need to define which parts of the analysis you want to run ('run_PanelDoc', 'run_Filtering', 'run_per_gene', 'run_Combine_results', 'calling_algorithm', 'run_CNVPanelizer') and minimun median coverage for you data ('minimum_coverage'). This is an important numerical parameter, the rest can be run with default values.

In the component results folder 8 subfolders will be created as outputs. The most important one is the 'calls' folder where the final copy number ratios will be written, as well as possible outputs from additional filtering, combined results etc. Another important file is the 'Platform_basedata.csv'. The creation of this file is time consuming, so if you need to run the same set of samples several times, or if the pipeline crashed AFTER (marked by the printed message "Platform data prepared") creating this file you can simply fetch the file from the outputs and give it as the input parameter 'partition_platform'. This will save time since the program doesn't need to start from the beginning.

The pipeline is quite time consuming (can take from 10 minutes up to two weeks) and requires some memory (For a set of 10-15 samples with 500 regions ~40GB). Having fewer regions reduces the running time, however we do not recommend to run only one or two chromosomes at a time since large CNVs will affect the normalization if the set is too small. The pipeline have been tested wih the 'Tumor' pipeline, and have been used succesfully used with default parameters. The 'Blood' pipeline runs without errors, however no further analysis of the pipeline has been done. Creating gene graphics has not been tested. Most problems that might arise during execution is due to issues with the input files, so please make sure to follow the instructions and test your data with a smaller sample set at first. Make sure to always add the last slash to the paths. Since the program does not use a pool of normal samples, normal samples can be run with the cancer samples. In this case you can filter the cancer samples against the normal samples at the end, removing unreliable signal. This does, however, not remove any possible germline variants present in the normal samples. The filtering is only possible with the 'Tumor' pipeline, with 2 mclust_states and if the names of the normal samples match exactly the cancer samples apart from the extension "normal" instead of "cancer". It is recommended to also check the original calls when analysing the data since the filtering might not be reliable for any other set of data.

CNVPanelizer

This is originally a R-package from bioconductor (https://bioconductor.org/packages/release/bioc/html/CNVPanelizer.html, Oliveira C, Wolf T (2018). CNVPanelizer: Reliable CNV detection in targeted sequencing applications. R package version 1.12.0.). For CNVPanelizer you need a folder with cancer bam files, a folder with normal bam files (these do not need to be matched since the program uses a pool of normals), and a bedfile.

Checklist for using the component:

In the parameter description the parameters are numbered.

1. Define parameters 1-5. Which algorithm do you want to use, and how do you want to process the outputs?

2. Based on your choices in step one, create and define the needed inputs for parameters 6-20.

3. Define parameters 21-29. Is this the first run or do you already have some of the intermediate files?

4. Define parameters 30-50. Default parameters can probably be used.

Outputs

Parameters

Test cases