\name{genoCNV}
\alias{genoCNV}
\title{ Copy Number Variation }
\description{
  extract genotype and copy number calls for copy number variation, which are inheritable DNA polymorphisms and are observed in normal tissues
}
\usage{
genoCNV(snpNames, chr, pos, LRR, BAF, pBs, sampleID, 
  Para=NULL, fixPara=FALSE, cnv.only=NULL, estimate.pi.r=TRUE, 
  estimate.pi.b=FALSE, estimate.trans.m=FALSE, normLRR=TRUE, 
  outputSeg=TRUE, outputSNP=3, outputTag=sampleID, outputViterbi=FALSE, 
  Ds = c(1e6, 1e6, rep(1e5, 4)), 
  pBs.alpha=0.001, loh=FALSE, output.loh=FALSE, 
  min.tp=5e-5, max.diff=0.1, distThreshold=5000, 
  transB = c(0.995, 0.005*c(.01, .09, .8, .09, .01)), 
  epsilon=0.005, K=5, maxIt=200, seg.nSNP=3, traceIt=5)
}
\arguments{
  \item{snpNames}{ a vector of SNP names. SNPs must be ordered by chromosome locations }
  
  \item{chr}{ chromosomes of all the SNPs specified in \code{snpNames} }
  
  \item{pos}{ positions of all the SNPs specified in \code{snpNames}}
  
  \item{LRR}{ Log R Ratio of all the SNPs specified in \code{snpNames}}
  
  \item{BAF}{ B Allele Frequency of all the SNPs specified in \code{snpNames}}
  
  \item{pBs}{ population frequency of of all the SNPs specified in \code{snpNames} }

  \item{sampleID}{ symbol/name of the studied sample. Only one sample is studied each time }

  \item{Para}{ a list of initial parameters for the HMM. If Para is NULL, The 
   default initial parameters: init.Para.CNV is used }

  \item{fixPara}{ if fixPara is TRUE, the parameters in Para are fixed, and are used directly to calculate posterior probabilities }

  \item{cnv.only}{ a vector indicating those CNV-only probes, for which we 
   only consider their Log R ratio. If it is NULL, there is no CNV-only probes }

  \item{estimate.pi.r}{ to estimate pi.r (proportion of uniform component for LRR) or not. By default, estimate.pi.r=FALSE, and the initial value of pi.r is used to estimate other parameters }

  \item{estimate.pi.b}{ to estimate pi.b (proportion of uniform component for BAF) or not. By default, estimate.pi.b=FALSE, and the initial value of pi.b is used to estimate other parameters }
	
  \item{estimate.trans.m}{ to estimate transition probability matrix or not. By default, estimate.trans.m=FALSE, and the initial value of estimate.trans.m is used to estimate other parameters }

  \item{normLRR}{ If normLRR is TRUE, we normalize the LRR data by subtracting 
  the median LRR for those LRR between -2 and 2. This strategy has been used by
  PennCNV.}
	
  \item{outputSeg}{ wether to output the information of copy number altered segments }
  
  \item{outputSNP}{ if \code{outputSNP} is 0, do not output SNP specific information; if \code{outputSNP} is 1, output the most likely copy number and genotype state of the SNPs that are within copy number altered regions; if \code{outputSNP} is 2, output the most likely copy number and genotype state of all the SNPs (whether it is within CNV regions or not), if \code{outputSNP} is 3, output the posterior probability for all the copy number and genotype states for the SNPs. }
  
  \item{outputTag}{ the prefix of the output files, output of copy number altered segments is written into file outputTag\_segment.txt, and output of SNP information is written into file outputTag\_SNP.txt }
  
  \item{outputViterbi}{ whether to output the copy altered regions identified by the viterbi algorithm. see details }
  
  \item{Ds}{ Parameter to for transition probability of the HMM. A vector of length N, where N is the number of states in the HMM }
  
  \item{pBs.alpha}{
    pBs.alpha is the lower limit of population B allele frequency, and the 
    upper limit is 1 - pBs.alpha
  }
  
  \item{loh}{ Whether we use the copy-number-neutral loss of heterozygosity state for CNV studies.  }

  \item{output.loh}{ Whether we output the loh information.  }
  
  \item{min.tp}{ the minimum of transition probability. }
  
  \item{max.diff}{ Due to normalization procedure, the BAF may not be symmetric. Let's use state (AAA, AAB, ABB, BBB) as an example. Ideally, mean values of normal components AAB and ABB, denoted by mu1 and mu2, respectively, should have the relation mu1 = 1-mu2 if BAF is symmetric. However, this may not be true due to normalization procedures. We restrict the difference of mu1 and (1-mu2) by this parameter max.diff. }

  \item{distThreshold}{ If distance between adjacent probes is larger than 
  distThreshold, restart the transition probability by the default values
  in \code{transB}.}

  \item{transB}{ The default transition probability. }
    
  \item{epsilon}{ see explanation of \code{K} }

  \item{K}{ epsilon and K are used to specify the convergence
  criteria. We say the estimate.para is converged if for K consecutive 
  updates, the maximum change of parameter estimates in every adjacent
  step is smaller than epsilon }
  
  \item{maxIt}{
    the maximum number of iterations of the EM algorithm to estimate parameters
  }
  
  \item{seg.nSNP}{
    the minimum number of SNPs per segment
  }
  
  \item{traceIt}{
    if traceIt is a integer n, then the running time is printed out in every n iterations of the EM algorithm.
    if traceIt is 0 or negative, no tracing information is printed out.
  }
}
\value{
  results are written into output files
}
\note{
  Copy number altered regions are identified, by default, based on the SNP level copy number calls. A CNV region boundary is declared simply when the adjacent SNPs have different copy numbers. An alternative approach is to use viterbi algorithm to output the ``best path''. Most time the results based on the SNP level copy number calls are the same as the results from viterbi algorithm. For the following up association studies, the SNP level information is more relevant if we examine the association SNP by SNP. 
}
\examples{
data(snpData)
data(snpInfo)

dim(snpData)
dim(snpInfo)

snpData[1:2,]
snpInfo[1:2,]

snpInfo[c(1001,1100,10001,10200),]

plotCN(pos=snpInfo$Position, LRR=snpData$LRR, BAF=snpData$BAF, 
main = "simulated data on Chr22")

snpNames = snpInfo$Name
chr = snpInfo$Chr
pos = snpInfo$Position
LRR = snpData$LRR
BAF = snpData$BAF
pBs = snpInfo$PFB
cnv.only=(snpInfo$PFB>1)
sampleID="simu1"

Theta = genoCNV(snpNames, chr, pos, LRR, BAF, pBs, 
            sampleID, cnv.only=cnv.only, outputSeg = TRUE, 
            outputSNP = 1, outputTag = "simu1")
}
\author{ Wei Sun and Zhengzheng Tang }
\keyword{ methods }