search

aeeckhou / shallowHRD

This method uses shallow Whole Genome Sequencing (sWGS) and the segmentation of a genomic profile to assess the Homologous Recombination Deficiency of a tumor based on the number of Large-scale Genomic Alterations (LGAs).
30 stars 13 forks source link

readme

shallowHRD

This method uses shallow Whole Genome Sequencing (sWGS > 0.3x) and the segmentation of a tumor genomic profile to infer the Homologous Recombination status of a breast and ovarian tumor based on the number of Large-scale Genomic Alterations (LGAs), evaluated in a similar way to LSTs (Large-scale State Transitions) but independent of the ploidy, with no reference to an absolute copy number. This can also be applied to pancreatic and prostate tumor.

Introduction

shallowHRD is a R script that can be launched from the command line. It relies on a ratio file characterizing the normalized read counts of a shallow Whole Genome Sequencing (>0.3x) in sliding windows along the genome and its segmentation. It was developped on the output of ControlFREEC (Boeva,V. et al., 2012) but is adapted to similar softwares. We recommand now to use QDNAseq with 50kb windows (see QDNAseq_script_chrX). A script is also provided for ControlFREEC output. Adaptation to other tools are however possible by matching the required input format (see sections "run shallowHRD" and "Nota Bene").

Softwares such as QDNAseq count reads in sliding windows, normalize read count and then segment the genomic profile. shallowHRD, based on a inferred CNA cut-off representing a one copy difference, will smooth the segmentation in a step wise manner, using first large segments, reintegrating small segments afterwards and then filtering small interstitial CNAs. The profile is optimised two times for a more robust output and the inferred CNA cut-off is each time based on simulations. The HR status is estimated based on the number of Large-scale Genomic Alterations (LGAs) i.e. intra-chromosome arm CNA breaks along the genome.

IMPORTANT : This GitHub contains the first version of shallowHRD (v1.13). Since its publication, the software has been under continuous developpement and the shallowHRDv2 has been published in Oncogene on November 2023. It improves shallowHRD by (i) securing correct estimation of LGA by managing specific noise coming from FFPE samples and (ii) minimizing not conclusive diagnostics by resolving borderline cases. It has been validated against the PAOLA-1/ENGOT-OV25 phase-III trial. The version 2.0 is for now not available online.

Requirements

Tested on Linux, Mac and Windows.

Prerequisities

First, FASTQ files should be aligned to a reference genome (hg19 or hg38) (using BWA-MEM for instance) and supplementary & duplicate reads removed from the BAM files, using Samtools and PicardTools' MarkDuplicates, respectively.

IMPORTANT: Please only use chromosomes 1 to 22 (plus the Chromosome X if you want to) for the alignment step. Additionnal chromosomes (contigs) might introduce errors.

Then, the BAM file should then be processed by a software such as ControlFREEC. The recommended options for controlFREEC are indicated in a config file example in the repository (controlfreec_config_file_example_hg19.txt). The window size was fixed here to 20kb (coverage > 0.4x) and the parameters were set for a sensitive segmentation. The window size can however be increased up to ~60kb if necessary depending on the coverage, with a step size half its length.

Finally, the file cytoBand_adapted_hg19.csv or cytoBand_adapted_hg38.csv (available in the repository) has to be downloaded.

The R packages needed can be installed with the script install_packages.R (in repository) and the command line :

/path/to/Rscript /path/to/install.packages.R

Run shallowHRD

To run shallowHRD only one ratio file is needed (formated in ControlFREEC's output).

The name of the file should be in this format : SAMPLE_NAME.bam_ratio.txt.

shallowHRD will rely on the first four columns of the input file (tabulated and with column Chromosome in number) :
Chromosome   Start   Ratio   RatioMedian
1    1     -1     -1
1    20001    -1    -1
.    .    .    .
.    .    .    .

The command line to launch shallowHRD is (absolute or relative paths) :

/path/to/Rscript /path/to/shallowHRD_hg19.R /path/to/SAMPLE_NAME.bam_ratio.txt /path/to/output_directory /path/to/cytoBand_adapted_hg19.csv

For Windows, it will be with /path/to/Rscript.exe.

Two examples in hg19 and one example in hg38 are downloadable in the repository to try shallowHRD.

Outputs

All the figures and files created by the script will be available in the output directory.

The summary plot figure recapitulating all the information will look like this :

alt text

A : Genomic profile with LGAs in green (the entire processed segmentation is represented in red if there are no LGA)
B : Density representing pairwise comparison between large segments used to fix the difference for a copy level
C : Graphe representing the value of each final segment (small blue circles) -
       If the segmentation is good, the different copy number should appear clearly with disctinct steps
D : Table recapitulating different data, including the case quality and the final diagnostic for the HR status

Nota Bene

  1. The scripts for QDNAseq and controlfreec have been updated to the 1.13 version. They harbor a more robust CNA cut-off detection and overall optimization of the profiles

  2. The 1.13 version of shallowHRD is more robust and reliable but takes a longer time to run compared to older version (~1 hour by sample)

  3. Different scripts for QDNAseq and controlfreec are available depending on whether the chromosome X is included in the ratio file

  4. shallowHRD can be adapted to other softwares with slight modification of outputs to match shallowHRD intput format

  5. The overall pipeline works also on WGS with a higher coverage

Contact

Don't hesitate to contact us for any questions, problems or adaptation of the method !

eeckhoutte.alexandre@gmail.com
tatiana.popova@curie.fr
marc-henri.stern@curie.fr

Publications

shallowHRD publication :

Alexandre Eeckhoutte, Alexandre Houy, Elodie Manié, Manon Reverdy, Ivan Bièche, Elisabetta Marangoni, Oumou Goundiam, Anne Vincent-Salomon, Dominique Stoppa-Lyonnet, François-Clément Bidard, Marc-Henri Stern, Tatiana Popova. ShallowHRD: Detection of Homologous Recombination Deficiency from shallow Whole Genome Sequencing. Bioinformatics (2020), https://doi.org/10.1093/bioinformatics/btaa261

shallowHRDv2 publication :

Celine Callens, Manuel Rodrigues, Adrien Briaux, Eleonore Frouin, Alexandre Eeckhoutte, Eric Pujade-Lauraine, Victor Renault, Dominique Stoppa-Lyonnet, Ivan Bieche, Guillaume Bataillon, Lucie Karayan-Tapon, Tristan Rochelle, Florian Heitz, Sabrina Chiara Cecere, Maria Jesús Rubio Pérez, Christoph Grimm, Trine Jakobi Nøttrup, Nicoletta Colombo, Ignace Vergote, Kan Yonemori, Isabelle Ray-Coquard, Marc-Henri Stern & Tatiana Popova. Shallow whole genome sequencing approach to detect Homologous Recombination Deficiency in the PAOLA-1/ENGOT-OV25 phase-III trial. Oncogene (2023), https://doi.org/10.1038/s41388-023-02839-8