CNsignatures

If you wish to generate copy number signature activities / exposures you can do so using the CINSignatureQuantification R package which supports both the new pan-cancer copy number signatures and the ovarian copy number signatures (ones described in this repository).

This repository contains all code and documentation necessary to reproduce the analysis undertaken in the manuscript Copy-number signatures and mutational processes in ovarian carcinoma

Source code

The source code for copy-number signature analysis is split across two files:

• main_functions.R contains a set of wrapper functions which allow identification and quantification of copy-number signatures.
• helper_functions.R contains the underlying functions that carry out the copy-number signature analysis.

These functions require the following packages to be installed:
NMF
flexmix
QDNAseq
YAPSA

A brief description of the wrapper functions:

• extractCopynumberFeatures This function takes as input a collection of absolute copy-number profiles and returns a list of copy-number features extracted from these samples. Copy-number profiles can be input as either a QDNAseq object, or as a list of segment tables. The segment tables (one for each sample) should have the following column headers: "chromosome", "start", "end", "segVal".
• fitMixtureModels This function takes as input a list of copy nnumber features and outputs a list of mixture model components fit to these features.
• generateSampleByComponentMatrix Given a set of extracted copy number features profiles this function returns a sum-of-posterior sample-by-component matrix. If the all_components argument is specified, then the sum-of-posteriors is calculated using these components, otherwise the component definitions from the manuscript are used.
• chooseNumberSignatures This function provides a wrapper to nmf functions required for determining the number of copy-number signatures in a dataset. It takes as input sample-by-component matrix and outputs a diagnostic plot for determining the optimal number of copy-number signatures.
• generateSignatures This function takes a sample-by-component matrix and a specified number of signatures and provides a wrapper to the nmf function for identifying copy-number signatures which returns a NMFfit object.
• quantifySignatures Given a sample-by-component matrix this function quantifies signature exposures using the LCD function from the YAPSA package, returning a normalised signature-by-sample matrix. If the component_by_signature matrix is specified then this matrix is used to define the signatures otherwise the signature definitions from the manuscript are used.

Data files

The following data files, located in data/ are required for copy-number signature analysis:

• __component_parameters.rds__ A RDS object containing the mixture model compenent definitions from the manuscript.
• feat_sig_mat.rds A RDS object containing the signature definition matrix from the manuscript.
• __gap_hg19.txt__ An annotation file containing the centromere locations for human hg19 genome build.
• hg19.chrom.sizes.txt An annotation file for the hg19 build chromosome lengths.

How to quantify existing signatures in your own samples

Once you have formatted your copy number data into a list of segment tables, run the following functions:

• CN_features<-extractCopynumberFeatures(<segment tables>)
• sample_by_component<-generateSampleByComponentMatrix(CN_features)
• quantifySignatures(sample_by_component)

How to generate your own signatures (advanced user)

Once you have formatted your copy number data into a list of segment tables, run the following functions:

• CN_features<-extractCopynumberFeatures(<segment tables>)
• all_components<-fitMixtureModels(CN_features) (this step is optional)
• sample_by_component<-generateSampleByComponentMatrix(CN_features,all_components)
• chooseNumberSignatures(sample_by_component)
• component_by_signature<-generateSignatures(sample_by_component,<num signatures>)
• quantifySignatures(sample_by_component,component_by_signature)

Manuscript analysis Rmarkdown

The manuscript_Rmarkdown/ directory contains a collection of markdown documents which when compiled, reproduce the entire analysis carried out in the manuscript. To compile this document run knitr to html on the CN_signature_main.R file.

Packages required to compile the Rmarkdown document include: knitr RColorBrewer ggplot2 plyr dplyr QDNAseq Biobase survival survcomp flexmix DOSE ppcor VariantAnnotation org.Hs.eg.db TxDb.Hsapiens.UCSC.hg19.knownGene rtracklayer ReactomePA Hmisc corrplot ggpubr NbClust kableExtra