postPING: Post process Estimation of binding site positions obtained...

In SRenan/PING: Probabilistic inference for Nucleosome Positioning with MNase-based or Sonicated Short-read Data

Description

Post process Estimation of binding site positions obtained from PING. Refit mixture models with stronger prior in candidate regions contain potential problems, and then convert final result into dataframe.

Usage

postPING(
  ping,
  seg,
  rho2 = NULL,
  sigmaB2 = NULL,
  alpha2 = NULL,
  beta2 = NULL,
  min.dist = 100,
  paraEM = NULL,
  paraPrior = NULL,
  score = 0.05,
  dataType = "MNase",
  nCores = 1,
  makePlot = FALSE,
  FragmentLength = 100,
  mart = NULL,
  seg.boundary = NULL,
  DupBound = NULL,
  IP = NULL,
  datname = ""
)

Arguments

ping	A pingList object containing estimation of nucleosome positions as returned by the PING function.
seg	An object of class segmentReadsList containing the results for all pre-processed regions as returned by segmentReads.
rho2, sigmaB2, alpha2, beta2	Integer values, the parameters in the prior of mixture models to be re-fitted.
min.dist	The minimum distance of two adjacent nucleosomes predicted from different candidate regions, smaller than that will be treated as duplicated predictions for the same nucleosomes.
paraEM	A list of parameters for the EM algorithm. The default parameters should be good enough for most usages.
paraPrior	A list of parameters for the prior distribution. The default parameters should be good enough for most usages.
score	A numeric. The score threshold used when calling FilterPING.
dataType	A character that can be set to use selected default parameters for the algorithm.
nCores	An integer. The number of cores that should be used in parallel by the function.
makePlot	A logical. Plot a summary of the output.
FragmentLength	An integer. The length of XSET profile extension
mart, seg.boundary, DupBound, datname	Plotting parameters and options.
IP	A GRanges object. The reads used in segmentation process.
minK	An integer.The minimum number of binding events per region. If the value is 0, the minimum number is automatically calculated.
maxK	An integer. The maximum number of binding events per region. If the value is 0, the maximum number is automatically calculated.
tol	A numeric. The tolerance for the EM algorithm.
B	An integer. The maximum number of iterations to be used.
mSelect	A character specifying the information criteria to be used when selecting the number of binding events. Default="AIC3"
mergePeaks	A logical stating whether overlapping binding events should be picked.
mapCorrect	A logical stating whether mappability profiles should be incorporated in the estimation, i.e: missing reads estimated.
xi	An integer. The average DNA fragment size.
rho	An integer. A variance parameter for the average DNA fragment size distribution.
alpha	An integer. First hyperparameter of the inverse Gamma distribution for sigma^2 in the PICS model
beta	An integer. Second hyperparameter of the inverse Gamma distribution for sigma^2 in the PING model
lambda	An integer. The lambda control Gaussian Markov Random Field prior on the distance of adjacent nucleosomes, we do not recommend user change the default value.
dMu	An integer. Our best guess for the distance between two neighboring nucleosomes.

Value

A data.frame containing the estimated binding site positions

Note

Based on our experiemt on a few real data sets, we suggestion to use following values of parameters. For sonication data we use rho1=1.2; sigmaB2=6400; rho=15; alpha1=10; alpha2=98; beta2=200000. For MNase data we use rho1=3; sigmaB2=4900; rho=8; alpha1=20; alpha2=100; beta2=100000. The value of xi depends on specs of sample, since that affect the length of linker-DNA. For example, we use xi=160 for yeast and xi=200 for mouse.

See Also

PING, plotSummary

SRenan/PING documentation built on Dec. 31, 2019, 12:02 p.m.