make_pen: =========================================================================...

In CyanolabFreiburg/rifi: 'rifi' analyses data from rifampicin time series created by microarray or RNAseq

========================================================================= make_pen

make_pen assigns automatically a penalties

Description

'make_pen' calls one of four available penalty functions to automatically assign penalties for the dynamic programming.

Usage

make_pen(
  inp,
  FUN,
  cores = 1,
  logs,
  dpt = 1,
  smpl_min = 10,
  smpl_max = 100,
  sta_pen = 0.5,
  end_pen = 4.5,
  rez_pen = 9,
  sta_pen_out = 0.5,
  end_pen_out = 3.5,
  rez_pen_out = 7
)

Arguments

inp	SummarizedExperiment: the input data frame with correct format.
FUN	function: one of the four bottom level functions (see details)
cores	integer: the number of assigned cores for the task
logs	numeric vector: the logbook vector.
dpt	integer: the number of times a full iteration cycle is repeated with a more narrow range based on the previous cycle. Default is 2.
smpl_min	integer: the smaller end of the sampling size. Default is 10.
smpl_max	integer: the larger end of the sampling size. Default is 100.
sta_pen	numeric: the lower starting penalty. Default is 0.5.
end_pen	numeric: the higher starting penalty. Default is 4.5.
rez_pen	numeric: the number of penalties iterated within the penalty range. Default is 9.
sta_pen_out	numeric: the lower starting outlier penalty. Default is 0.5.
end_pen_out	numeric: the higher starting outlier penalty. Default is 3.5.
rez_pen_out	numeric: the number of outlier penalties iterated within the outlier penalty range. Default is 7.

Details

The four functions to be called are:

1. fragment_delay_pen

2. fragment_HL_pen

3. fragment_inty_pen

4. fragment_TI_pen

These functions return the amount of statistically correct and statistically wrong splits at a specific pair of penalties. 'make_pen' iterates over many penalty pairs and picks the most suitable pair based on the difference between wrong and correct splits. The sample size, penalty range and resolution as well as the number of cycles can be customized. The primary start parameters create a matrix with n = rez_pen rows and n = rez_pen_out columns with values between sta_pen/sta_pen_out and end_pen/end_pen_out. The best penalty pair is picked. If dept is bigger than 1 the same process is repeated with a new matrix of the same size based on the result of the previous cycle. Only position segments with length within the sample size range are considered for the penalties to increase run time. Returns a penalty object (list of 4 objects) the first being the logbook.

Value

A list with 4 items:

logbook:

Interger, the logbook vector containing all penalty information

penalties:

Integer, a vetor with the respective penalty and outlier penalty

correct:

Matrix, a matrix of the correct splits

wrong:

Matrix, a matrix of the incorrect splits

Examples

data(fit_minimal)
make_pen(
  inp = fit_minimal, FUN = rifi:::fragment_HL_pen, cores = 2,
  logs = as.numeric(rep(NA, 8)), dpt = 1, smpl_min = 10, smpl_max = 50,
  sta_pen = 0.5, end_pen = 4.5, rez_pen = 9, sta_pen_out = 0.5,
  end_pen_out = 3.5, rez_pen_out = 7
)

CyanolabFreiburg/rifi documentation built on May 7, 2023, 7:53 p.m.