version 1.0

DESCRIPTION

@@ -0,0 +1,19 @@
+Package: enRich
+Type: Package
+Title: An R package for analysis of multiple ChIP-seq data
+Version: 1.0
+Date: 2013-05-22
+Author: Yanchun Bao <yanchun.bao@brunel.ac.uk>, Veronica
+        Vinciotti<veronica.vinciotti@brunel.ac.uk>
+Maintainer: Yanchun Bao <yanchun.bao@brunel.ac.uk>
+Description: enRich is an R package for joint statistical modelling of
+        ChIP-seq data, accounting for technical/biological replicates,
+        multiple conditions and different ChIP efficiencies of the
+        individual experiments.
+Depends: R(>= 2.14), parallel
+License: GPL (>= 2)
+LazyLoad: yes
+Packaged: 2013-05-31 12:00:04 UTC; masryyb
+NeedsCompilation: no
+Repository: CRAN
+Date/Publication: 2013-05-31 14:22:44

MD5

@@ -0,0 +1,24 @@
+73191c776a0d570e7d09da07f3b57729 *DESCRIPTION
+8ccb633701f57d5020c962e5a68931ab *NAMESPACE
+c2d161145fe5f54aaf850dd0b86b24b7 *R/FDR.R
+0de35362a72a63dae57766d58329d476 *R/IPE.R
+b9b79fd941de0876c76cfeefe19deac1 *R/NBlike_phi_offset.R
+1e36698d0ceaf5bfe3174e2267056c9b *R/enrich.mix.R
+a8efc832733c327427778a866133bb17 *R/mix.R
+b6b15cddafa3097b4c1dbc74d5a11edf *R/mix.joint.R
+00f90607b7e2d8aab26eeaeeae286d40 *R/mix_offset.R
+3e673636f87af251956b7b63ee1b5a16 *R/mix_single.R
+041321ec3173ac886ef5efe78023bbff *R/mix_srsp1.R
+00f90607b7e2d8aab26eeaeeae286d40 *R/mixfit_offset.R
+257d3aaf3ebe645a537cc452c0d2c639 *R/mloglike_offsetNB.R
+cb8c5f8392eb06ee89f5df9750d90fac *R/mloglike_offsetpois.R
+9cb4b5212ff81b46d91c07733506e438 *R/pprob.R
+76e34f89608959cbe3c821ed057edaf0 *R/pprob_joint.R
+91637870bd91aa78972be5d2e3e937e9 *data/p300cbp.1000bp.rda
+b4a38667ca28859ecb5a237bd01ec1c5 *man/FDR.Rd
+0dd607094201c7fe4e65b060045af624 *man/IPE.Rd
+35b21f3f1aa099551c359db4b149c89f *man/enRich-package.Rd
+8a6580e5cce23f7580da332410d93852 *man/enrich.mix.Rd
+28f3499c42b1c79e62b739a03ceb1d84 *man/mix.Rd
+78b118a4cb8bd7f631349bc4d81d9a1c *man/mix.joint.Rd
+494ad59e5cb59d8c588bde57d032c170 *man/p300cbp.1000bp.Rd

NAMESPACE

@@ -0,0 +1,2 @@
+export(enrich.mix, FDR, IPE, mix.joint, mix)
+

R/FDR.R

@@ -0,0 +1,26 @@
+FDR <-
+function(prob0, cr=0.05)
+{
+## prob0 is the posterior probability of X=0
+   S=length(prob0)
+   Pvalue=prob0
+   sortPvalue=Pvalue[order(Pvalue)]
+   positionp2=order(Pvalue)
+   newPvalue=rep(0, S)
+   positionp1=order(Pvalue)
+   tempp=sortPvalue[1]
+   newPvalue[1]=tempp/1
+   for (s in 2:S)
+   {
+      tempp=tempp+sortPvalue[s]
+      newPvalue[s]=tempp/s
+   }
+   newPvalueback=newPvalue[order(positionp1)]
+   tempX<-ifelse(newPvalueback<=cr, 1,0)
+   cpoint=sum(tempX)
+   T_BH=ifelse(cpoint==0, 0, sortPvalue[cpoint]) #Threshold
+   tempX<-ifelse(Pvalue<=T_BH, 1, 0) ## Consider the tie
+   FDR_t=ifelse(sum(tempX)==0, 0, sum(Pvalue[Pvalue<=T_BH])/sum(tempX))
+   result<-list(X=tempX)
+   return(result)
+}

R/IPE.R

@@ -0,0 +1,48 @@
+IPE <-function(para, method=NULL)
+{
+## para is the fitting results of one experiment
+## method is "Poisson" or "NB"
+
+	#1. Check data, method and initialpara are consistent
+	if (sum(method=="Poisson")+sum(method=="NB")==0)
+	{
+		stop('A method must be given, either "Poisson" or "NB"!', call.=FALSE)
+	}
+
+	## we need to use interpolation since the discrete distribution will give some larger mass on some particular value
+	## we interpolate between all values below cut
+	T=10000
+	cut=100
+	n=500
+	sum1=0
+	sum2=0
+	if (method=="Poisson")
+	K=4
+	if (method=="NB")
+	K=6
+	for (i in para[K]:cut)
+	{
+		if (method=="Poisson")
+		{
+			X.bg<-approx(x=c(i, i+1),y=c(dpois(i,para[3]),dpois(i+1,para[3])), n=n)
+			Step<-X.bg$x[2]- X.bg$x[1]
+			sum2=sum2+sum(ppois(i-para[K], para[2])*X.bg$y)*Step 
+		}
+		if (method=="NB")
+		{
+			X.bg<-approx(x=c(i, i+1),y=c(dnbinom(i,para[5], ,para[4]),dnbinom(i+1,para[5], ,para[4])), n=n)
+			Step<-X.bg$x[2]- X.bg$x[1]
+			sum2=sum2+sum(pnbinom(i-para[K], para[3],,para[2])*X.bg$y)*Step 
+		}
+	}
+	if (method=="Poisson")
+	{
+		sum2=sum2+ sum(ppois(c((cut+1-para[K]):(T-para[K])), para[3])*dpois(c((cut+1):T), para[2]))
+	}
+	if (method=="NB")
+	{
+		sum2=sum2+ sum(pnbinom(c((cut+1-para[K]):(T-para[K])), para[3],, para[2])*dnbinom(c((cut+1):T), para[5],,para[4]))
+	}
+	IPE=round(1-sum2, digits=4)
+	return(IPE)
+}

R/NBlike_phi_offset.R

@@ -0,0 +1,12 @@
+NBlike_phi_offset <-
+function(para, data1, ppr1, ppr0, mus,mub, k=k)
+{
+#Log-likelihood based on mixture of NB(mu, size), maximised using optim() function
+   N=length(data)
+   phis=para[1]
+   phib=para[2]
+   temp1=sum(ppr1*ifelse(data1<k, 0, dnbinom(data1-k, phis, ,mus, log=TRUE)))
+   temp2=sum(ppr0*dnbinom(data1, phib, ,mub, log=TRUE))
+   logl=temp1+temp2
+   return(-logl)
+}