pepBayesEcm-methods: Posterior maximization for pepBayes models.
In RGLab/pepBayes: Bayesian modeling of peptide microarray assay responses

Description Usage Arguments Details Value See Also

Maximize the posterior distribution of pepBayes models. The generic function has methods for peptideSet objects and for pepBayesFit objects for restarting an optimization from its last position.

pepBayesEcm(x, ...)

## S4 method for signature 'peptideSet'
pepBayesEcm(x, position_data = NULL,
  control_id = NULL, n_iter = 30, n_threads = 1, n_rule = 10,
  n_isamp = 20, prior_control_list = NULL, iter_control_list = NULL,
  schedule_iterations = TRUE)

## S4 method for signature 'pepBayesFit'
pepBayesEcm(x, n_iter, n_threads = 1, n_rule = 10,
  n_isamp = 20, schedule_iterations = FALSE, join_output = TRUE)

`x`	A pepBayesFit or peptideSet object containing experiment data.
`...`	Additional arguments for sampling control.
`position_data`	A data.frame or GRanges object with peptide position information.
`n_iter`	The number of EM iterations to perform.
`n_threads`	The number of parallel threads to use in the E-step.
`n_rule`	The number of nodes to use for Gauss-Hermite numerical integration.
`n_isamp`	The number of importance samples to estimate the E-step.
`prior_control_list`	A named list controlling initial parameter values.
`iter_control_list`	A named list controlling which parameters are updated.
`schedule_iterations`	A boolean. Should the number of importance samples gradually increase to `n_isamp` as iterations progress?
`join_output`	A boolean. Should input be merged with output?

PepBayes estimates the posterior probability of a differential antibody response against each peptide in a peptide microarray assay, for each subject, when compared with control samples. Two common experimental designs are automatically detected. In an unpaired design, differences between a population of interest and a control population are detected. A paired design has matched samples for each subject before and after a treatment is applied. The function pepBayesEcm automatically detects the experimental design from the data.

When x is a peptideSet object, the paired or unpaired model is chosen depending on the 'ptid' column of pData(x). Paired data is detected when each unique value of 'ptid' appears exactly twice among all slides.

The position_data argument supplies information about peptide position information. This argument may be a 'data.frame' with columns 'start', 'end' or 'width', and 'peptide'. If a 'GRanges' object, then it must have either peptide as a name or have peptide as a metadata column. Response probability hyperparamters are fixed for EM estimation, but if position data are omitted (position_data = NULL), they will also be fixed if the resulting fitted object is used to initialize a pepBayesMcmc.

When x is a pepBayesFit object, the optimization is started from the most recent parameter values in the fitted object. In the event that x is the output of a previous ECM fit, the argument join_output becomes relevant. If TRUE, parameter traces are appended to those in x.

An object of type pepBayesFit containing sorted slide metadata, parameter trace, posterior probabilities of binding, and iteration information.

pepBayesFit pepBayesFit-methods peptideSet pepBayesMcmc pepBayesCalls