cubfits (version 0.1-2)

CUB Model Prediction: Codon Usage Bias Prediction for Observed ORFs

Description

This function provides codon usage bias fits of training set which has observed ORFs and expressions possibly containing measurement errors, and provides predictions of testing set which has other observed ORFs but without expression.

Usage

cubpred(reu13.df.obs, phi.Obs, y, n, reu13.df.pred, y.pred, n.pred, nIter = 1000, b.Init = NULL, init.b.Scale = .CF.CONF$init.b.Scale, b.DrawScale = .CF.CONF$b.DrawScale, b.RInit = NULL, p.Init = NULL, p.nclass = .CF.CONF$p.nclass, p.DrawScale = .CF.CONF$p.DrawScale, phi.Init = NULL, init.phi.Scale = .CF.CONF$init.phi.Scale, phi.DrawScale = .CF.CONF$phi.DrawScale, phi.pred.Init = NULL, phi.pred.DrawScale = .CF.CONF$phi.pred.DrawScale, model = .CF.CT$model[1], model.Phi = .CF.CT$model.Phi[1], adaptive = .CF.CT$adaptive[1], verbose = .CF.DP$verbose, iterThin = .CF.DP$iterThin, report = .CF.DP$report)

Arguments

reu13.df.obs

a reu13.df to be trained.

phi.Obs

a phi.Obs to be trained.

y

a y to be trained.

n

a n to be trained.

reu13.df.pred

a reu13.df to be predicted.

y.pred

a y to be predicted.

n.pred

a n to be predicted.

nIter

number of iterations after burn-in iterations.

b.Init

initial values for parameters b.

init.b.Scale

for initial b if b.Init = NULL.

b.DrawScale

scaling factor for adaptive MCMC with random walks when drawing new b.

b.RInit

initial values (in a list) for R matrices of parameters b yielding from QR decomposition of vglm() for the variance-covariance matrix of b.

p.Init

initial values for hyper-parameters.

p.nclass

number of components for model.Phi = "logmixture".

p.DrawScale

scaling factor for adaptive MCMC with random walks when drawing new sigma.Phi.

phi.Init

initial values for Phi.

init.phi.Scale

for initial phi if phi.Init = NULL.

phi.DrawScale

scaling factor for adaptive MCMC with random walks when drawing new Phi.

phi.pred.Init

initial values for Phi of predicted set.

phi.pred.DrawScale

as phi.DrawScale but for predicted set.

model

model to be fitted, currently "roc" only.

model.Phi

prior model for Phi, currently "lognormal".

adaptive

adaptive method of MCMC for proposing new b and Phi.

verbose

print iteration messages.

iterThin

thinning iterations.

report

number of iterations to report more information.

Value

A list contains four big lists of MCMC traces including: b.Mat for mutation and selection coefficients of b, p.Mat for hyper-parameters, phi.Mat for expected expression values Phi, and phi.pred.Mat for predictive expression values Phi. All lists have nIter / iterThin + 1 elements, and each element contains the output of each iteration.All lists also can be binded as trace matrices, such as via do.call("rbind", b.Mat) yielding a matrix of dimension number of iterations by number of parameters. Then, those traces can be analyzed further via other MCMC packages such as coda.

Details

This function correctly and carefully implements an extension of Shah and Gilchrist (2011) and Wallace et al. (2013).

Total number of MCMC iterations is nIter + 1, but the outputs may be thinned to nIter / iterThin + 1 iterations.

Temporary result dumping may be controlled by .CF.DP.

References

https://github.com/snoweye/cubfits/

Shah P. and Gilchrist M.A. ``Explaining complex codon usage patterns with selection for translational efficiency, mutation bias, and genetic drift'' Proc Natl Acad Sci USA (2011) 108:10231--10236.

See Also

DataIO, DataConverting, cubfits() and cubappr().

Examples

## Not run: 
# suppressMessages(library(cubfits, quietly = TRUE))
# 
# demo(roc.pred, 'cubfits', ask = F, echo = F)
# ## End(Not run)