mutSigMapper: maps spectra to mutational signatures

In juliancandia/mutSigMapper: Mapping Spectra to Mutational Signatures Based on Shot-Noise Modeling

Description

This method maps spectra to mutational signatures based on a robust, non-parametric statistical approach. For each spectrum (provided as the number of mutations observed across channels, e.g. single-substitution trinucleotide contexts), non-negative least square regression procedures are used to calculate the contributions of each mutational signature. Shot noise (implemented as a Poisson or negative binomial process) generates an ensemble of random spectra that enables the assessment of statistical significance for those contributions. The resulting object can be plotted and further analyzed using built-in functions provided in the package.

Usage

mutSigMapper(spectra, sig=NULL, ref=c("cosmic_v2","cosmic_v3",
"cosmic_v3_exome","cosmic_v3.1","mutagen53"), 
method=c("MutationalPatterns","deconstructSigs"),
sig.bkg.adj=c("none","1/exome","1/genome","exome/genome",
"genome/exome","custom"), sig.bkg.adj.custom=NULL,
signature.cutoff=0.06, noise=c("poisson","neg.binom"), 
neg.binom.size=NULL, n_rdm=200,
save_obj=FALSE, dest_dir=getwd(), dest_dir_create=TRUE, 
dest_dir_create_recur=FALSE, dest_obj="mutSigMap.Robj")

Arguments

spectra	Input numerical matrix or data frame with channels as rows and samples as columns. Data must be absolute counts (i.e. not normalized spectra). Channel and sample labels should be provided as row and column names, respectively.
sig	Input numerical matrix or data frame with channels as rows and signatures as columns. Channel and signature labels should be provided as row and column names, respectively. Channel labels must agree with those of the input spectra.
ref	If a sig data frame is not provided, a reference built-in compendium can be chosen. Options are "cosmic_v2" (Cosmic v2, March 2015), "cosmic_v3" and "cosmic_v3_exome" (the genome and exome releases from Cosmic v3, May 2019), "cosmic_v3.1" (Cosmic v3.1, June 2020), and "mutagen53" (mutagen compendium by Kucab et al, Cell (2019)).
method	Non-negative least squares regression procedure. Options are "MutationalPatterns" (default) and "deconstructSigs".
sig.bkg.adj	Distribution of background counts to adjust the signatures to match the spectra data source. See Details section below.
sig.bkg.adj.custom	For sig.bkg.adj="custom", data frame to adjust signatures. It should be provided in the same format as datasets ref_bkg_exome and ref_bkg_exome.
signature.cutoff	Signature contributions (weights) must be above this threshold.
noise	Statistical noise model to generate random spectra. Options are "poisson" and "neg.binom".
neg.binom.size	For negative binomial noise, this is the size parameter. Must be a positive number. As the size is increased, the distribution's variance decreases. Size values much larger than the distribution's mean yield negative binomial distributions similar to Poisson.
n_rdm	Number of random spectra generated for each sample. Default is 200.
save_obj	Logical to save the mutSigMapper object. Default is FALSE.
dest_dir	Destination directory. Default is the working directory.
dest_dir_create	Creates destination directory if it does not exist already. Default is TRUE.
dest_dir_create_recur	Creates destination directory recursively if it does not exist already. Default is FALSE.
dest_obj	Name for output mutSigMapper object.

Details

Input spectra must be provided as unnormalized mutation counts. An input signature matrix or data frame may be provided, but the package offers a variety of built-in reference compendia. Two different built-in procedures for non-negative least squares regression are included. Signatures may be adjusted to match the spectra data source. Signatures based on human genome abundance (such as "cosmic_v2", "cosmic_v3", "cosmic_v3.1" and "mutagen53") used to analyze spectra extracted from whole genome sequencing (WGS) do not require adjustment; for spectra extracted from whole exome sequencing (WES), however, one needs to adjust genome-based signatures using sig.bkg.adj="exome/genome". Conversely, WES-based signatures (such as "cosmic_v3_exome") assessed against WGS spectra need to be adjusted using sig.bkg.adj="genome/exome". Other potential mismatches between sequencing sources of spectra and signatures may arise from analyzing targeted sequencing data, as well as when analyzing data from different organisms. A sig.bkg.adj="custom" option is provided to allow users to adjust signatures based on a custom data frame of scale factors across channels via sig.bkg.adj.custom.

Value

An object with S3 class "mutSigMapper".

spectra	Input data frame.
sig	Input data frame.
method	Input parameter.
tri.counts.method	Input parameter.
signature.cutoff	Input parameter.
noise	Input parameter.
neg.binom.size	Input parameter.
n_rdm	Input parameter.
weights	List of weights (i.e. signature contributions) for each input sample.
weights_rdm	List of random weight matrices (rows=random spectra, columns=signatures) for each input sample.
map_pval	Matrix of significance p-values (rows=signatures, columns=samples).
counts	For method="MutationalPatterns", list of counts contributed across signatures for each input sample. This is the unnormalized version of weights.
counts_rdm	For method="MutationalPatterns", list of random count matrices (rows=random spectra, columns=signatures) for each input sample. This is the unnormalized version of weights_rdm.

Author(s)

Julian Candia
Maintainer: Julian Candia [e-mail]

References

Rosenthal R et al. DeconstructSigs: delineating mutational processes in single tumors distinguishes DNA repair deficiencies and patterns of carcinoma evolution, Genome Biol. 17, 31 (2016).

Blokzijl F et al. MutationalPatterns: comprehensive genome-wide analysis of mutational processes, Genome Med. 10, 33 (2018).

COSMIC (Catalogue Of Somatic Mutations In Cancer), https://cancer.sanger.ac.uk/cosmic/signatures/index.tt

Kucab JE et al. A Compendium of Mutational Signatures of Environmental Agents, Cell 177, 821-836.e16 (2019).

See Also

ref_bkg_exome, ref_bkg_genome, plotSpectraCaterpillar, plotSpectraHeatmap, export

Examples

# Spectra from Blokzijl et al:
data(spectra)
map = mutSigMapper(spectra)

juliancandia/mutSigMapper documentation built on Oct. 19, 2020, 3:58 a.m.