signeR: signeR
In rvalieris/signeR: Empirical Bayesian approach to mutational signature discovery

signeR

R Documentation

signeR

Description

Generates the signatures.

Usage

    signeR(M, Mheader = TRUE, samples = "rows", Opport = NA,
        Oppheader = FALSE,  P = NA, fixedP = FALSE,
        nsig = NA, nlim = c(NA, NA),
        try_all = FALSE, BICsignificance = FALSE, critical_p = 0.05,
        ap = NA, bp = NA, ae = NA, be = NA,
        lp = NA, le = NA, var.ap = 10, var.ae = 10,
        start = "lee", testing_burn = 1000, testing_eval = 1000,
        main_burn = 10000, main_eval = 2000,
        estimate_hyper = FALSE, EMit_lim=100, EM_eval = 100,
        parallelization = "multisession")

Arguments

`M`	mutation counts matrix of samples x features.
`Mheader`	if M has colnames defined use TRUE, if FALSE a default order will be assumed.
`samples`	if the samples are row-wise or column-wise in M, default is "row".
`Opport`	context count matrix of samples x features in the target genome or region.
`Oppheader`	if Opport has header defined.
`P`	Previously known matrix of signatures. If provided, can be fixed along algorithm iterations or just used as an initial value (see next parameter)
`fixedP`	If TRUE, provided P matrix will be fixed along iterations.
`nsig`	number of signatures, which can be provided or estimated by the algorithm.
`nlim`	define an interval to search for the optimal number of signatures.
`try_all`	if TRUE, all possible values for nsig will be tested
`BICsignificance`	if TRUE, BICs will be considered different only if their distribution is significantly different. In case of ties in BICs comparison, signer will adopt the model with fewer signatures.
`critical_p`	level of significance for BICs distribution to be considered different
`ap`	shape parameter of the gamma distribution used to generate the entries of a matrix of rate parameters of the gamma distributions which generate signatures.
`bp`	rate parameter of the gamma distribution used to generate the entries of a matrix of rate parameters of the gamma distributions which generate signatures.
`ae`	shape parameter of the gamma distribution used to generate the entries of a matrix of rate parameters of the gamma distributions which generate exposures.
`be`	rate parameter of the gamma distribution used to generate the entries of a matrix of rate parameters of the gamma distributions which generate exposures.
`lp`	parameter of the exponential distribution used to generate the entries of a matrix of shape parameters of the gamma distributions which generate signatures.
`le`	parameter of the exponential distribution used to generate the entries of a matrix of shape parameters of the gamma distributions which generate exposures.
`var.ap`	variance of the gamma distribution used to generate proposals for shape parameters of signatures
`var.ae`	variance of the gamma distribution used to generate proposals for shape parameters of exposures
`start`	NMF algorithm used to generate initial values for signatures and exposures,options: "brunet","KL","lee","Frobenius","offset","nsNMF", "ls-nmf","pe-nmf","siNMF","snmf/r" or "snmf/l".
`testing_burn`	number of burning iterations of the Gibbs sampler used to estimate the number of signatures in data. Corresponds to R0 at Algorithm 1 on signeR paper.
`testing_eval`	number of iterations of the Gibbs sampler used to estimate the number of signatures in data. Corresponds to R2 at Algorithm 1 on signeR paper.
`EM_eval`	number of samples generated at each iteration of the EM algorithm. Corresponds to R1 at Algorithm 1 on signeR paper.
`main_burn`	number of burning iterations of the final Gibbs sampler.
`main_eval`	number of iterations of the final Gibbs sampler.
`estimate_hyper`	if TRUE, algorithm estimates optimal values of ap,bp,ae,be,lp,le. Start values can still be provided.
`EMit_lim`	limit of EM iterations for the estimation of hyper-hyperparameters ap,bp,ae,be,lp,le. Default is 100. Corresponds to U at Algorithm 1 on signeR paper.
`parallelization`	strategy of computation parallelization, see future::plan help

Value

signeR output is a list with the following items:

`Nsign`	selected number of signatures.
`tested_n`	array containing the numbers of signatures tested by the algorithm.
`Test_BICs`	list of measured BIC values when testing different numbers of signatures.
`Phat`	Estimated signatures, median of P samples.
`Ehat`	Estimated exposures, median of E samples.
`SignExposures`	SignExp object which contains the set of samples for the model parameters.
`Bics`	measured BIC values on the final run of the sampler.
`HyperParam`	evolution of estimated hyperparameters when testing different numbers of signatures.