---
title: "MPRA REPORT"
output: html_document
---


```{r opts,echo=FALSE}
.libPaths( c("/n/home06/agroff/R/x86_64-unknown-linux-gnu-library/3.3",.libPaths()))
library(knitr)
library(rmarkdown)
```


```{r setup,echo=FALSE,warning=FALSE,message=FALSE}
library(ggplot2)
library(reshape)
library(plyr)
library(Biobase)
library(stringr)
library(reshape2)
opts_chunk$set(echo=FALSE, message=FALSE, warning=FALSE,fig.height=8, fig.width=8,fig.path = paste(filename,"images/",sep="/"), dev=c('png', 'pdf'))
```


```{r loaddata, echo=FALSE}

data_counts_orig<-read.table("C2C12_lincmRNA_combined_Jan2017_hiseq.tab",header=TRUE,stringsAsFactors = FALSE)
annotation<-read.table("lincmRNA_pool_annotation_updated_promotersinloci.txt",header=TRUE,stringsAsFactors = FALSE)
locus_TSS<-read.table("lincmRNA_TSSs.txt",header=TRUE,stringsAsFactors = FALSE,sep="\t")

threshold<-10
```




```{r tagsperelement}
loci<-c("CMV","Cdkn1a","Crnde","Enc1_mm10_refGene","Fendrr","Foxf1","Irx5","Lincenc1","Lincp21","Morc2a","Peril","Sox2_mm10","Tug1")
barcodes_to_keep<-annotation[grep(paste(loci,collapse="|"),annotation$name),]
barcodes_to_keep<-barcodes_to_keep[grep("Lincp21_promoter_mm10_refGene_NR_036469_scrambled|Sox2DistalSuperEnhancer|Cdkn1a_promoter_scrambled|Ef1a|TK",barcodes_to_keep$name,invert=TRUE),"barcode"]

data_counts_orig<-data_counts_orig[data_counts_orig$barcodes %in% barcodes_to_keep,]
QC.dat.annot<-merge(annotation,data_counts_orig,by.x="barcode",by.y="barcodes")
QC.dat.annot_touse<-QC.dat.annot
data_counts_touse<-data_counts_orig

#remove barcodes with DNA representation below threshold
data_dna<-data_counts_touse[,grep("VEC",names(data_counts_touse))]
row.names(data_dna)<-data_counts_touse$barcodes
data_dna[data_dna<threshold]<-0
data_dna[data_dna>=threshold]<-1
data_dna$sum<-rowSums(data_dna)
data_dna<-data_dna[which(data_dna$sum>=3),] #if 3 or more samples agree that it passes thresh then keep 

data_counts_touse<-data_counts_touse[which(data_counts_touse$barcodes %in% row.names(data_dna)),]

QC.dat.binary<-QC.dat.annot_touse[,c("Element")]
binary<-QC.dat.annot_touse[,c(14:dim(QC.dat.annot_touse)[2])]
binary[binary>0]<-1#if tags represented above threshold, call "1"
QC.dat.binary<-cbind(QC.dat.binary,binary)
names(QC.dat.binary)<-c("Element",names(binary))

tags<-aggregate(.~Element,data=QC.dat.binary,sum)

tags.annot<-merge(tags,QC.dat.annot_touse,by.x="Element",by.y="Element")

tags.melt<-melt(tags)
#ggplot(tags.melt,aes(variable,value))+geom_boxplot()+theme_bw()+theme(axis.text.x = element_text(angle = 65, hjust = 1))
```


```{r tagperrepdensity}
#Tag per element per replicate density plots: 
#ggplot(tags.melt,aes(value,fill=variable))+geom_density()+theme_bw()
#summary(tags)
```


```{r tagcountdist}
# tag distribution 
#distribution of counts
data_counts_touse.melt<-melt(data_counts_touse)
```


```{r norm}
# Assess replicate noise and tag noise within replicates
normalize_andcenter<-function(dat,threshold){
  dat2<-as.matrix(dat[,2:dim(dat)[2]])
  dat2[dat2<threshold]<-NA
  repsums<-colSums(dat2,na.rm=TRUE)
  dat_norm<-t(t(dat2)/repsums)
  medians<-apply(dat_norm,2,median,na.rm=TRUE)
  dat_norm2<-t(t(dat_norm)/medians)
  dat<-cbind(dat["barcodes"],dat_norm2)
  #namesfordat<-c("barcodes",names(dat_norm))
  as.data.frame(dat)
  #names(dat)<-namesfordat
  #dat
}
#dat_norm<-normalize(data_counts_touse,threshold)
dat_norm<-normalize_andcenter(data_counts_touse,threshold)
```


Histogram of normalized DNA and RNA samples: 

```{r histograms}
dat_norm.melt<-melt(dat_norm)

#DNA
DNA<-dat_norm.melt[grep("VEC",dat_norm.melt$variable),]
ggplot(DNA,aes(value))+geom_histogram()+theme_bw()
medDNA<-median(DNA$value,na.rm = TRUE)

#RNA
RNA<-dat_norm.melt[grep("VEC",dat_norm.melt$variable,invert=TRUE),]
ggplot(RNA,aes(value))+geom_histogram()+theme_bw()
medRNA<-median(RNA$value,na.rm = TRUE)

#median center (divide by median)

DNA$DNA_centered<-DNA$value/medDNA
RNA$RNA_centered<-RNA$value/medRNA

#DNA
ggplot(DNA,aes(DNA_centered))+geom_histogram()+theme_bw()+ggtitle("DNA centered")
median(DNA$DNA_centered,na.rm = TRUE)

#RNA
ggplot(RNA,aes(RNA_centered))+geom_histogram()+theme_bw()+ggtitle("RNA centered")
median(RNA$RNA_centered,na.rm = TRUE)

#median center dat_norm 
DNA_centered<-cast(DNA,barcodes~variable)
RNA_centered<-cast(RNA,barcodes~variable)

dat_norm_centered<-merge(RNA_centered,DNA_centered)
```


```{r scatters}
#replicate scatters/correlations
#Can deffinitely see some batch effects from RNA replicates from different passages (biological replicates). Dont see this with the Vec replicates. 

#library(GGally)
#pdf("pairs.pdf")
#ggpairs(dat_norm[,2:dim(dat_norm)[2]])
#dev.off()

library(stringr)
dat_norm2<-dat_norm_centered[,1:11]
colnames(dat_norm2)<-c("barcodes","C2C12_1","C2C12_2","C2C12_3","C2C12_4","C2C12_5","VEC_1","VEC_2","VEC_3","VEC_4","VEC_5")
dat_norm<-dat_norm2;rm(dat_norm2)
dat_norm.melt<-melt(dat_norm)
info<-str_split_fixed(dat_norm.melt$variable,"_",2)
dat_norm.melt$sampletype<-info[,1]
dat_norm.melt$replicate<-info[,2]
```



```{r signalprocessing}
# make signal ratios
#- average and variance across all samples
#- 100* randomly selected 2rna/3dna ratios
#- propagation of error for variance for each measurement?
#histograph of ratios (avRNA/avVEC) for this locus + controls (before shuffle): 
library(Biobase)

# ALL SAMPLES
#C2C12 vs Vec
#http://stackoverflow.com/questions/25099825/row-wise-variance-of-a-matrix-in-r
RowVar <- function(x) {
  ans<-sum((x - mean(x,na.rm=TRUE)^2),na.rm=TRUE)/(length(x) - 1)
  ans
}

#change name to eventually make function 
names(dat_norm)<-gsub("C2C12","RNA",names(dat_norm))
names(dat_norm)<-gsub("lincmRNAvec","VEC",names(dat_norm))

RNA<-dat_norm[,grep("RNA",names(dat_norm))]
row.names(RNA)<-dat_norm$barcodes
VEC<-dat_norm[,grep("VEC",names(dat_norm))]
row.names(VEC)<-dat_norm$barcodes

rnaAvg<-rowMedians(as.matrix(RNA))
rnaVars<-apply(RNA,1,RowVar)
vecAvg<-rowMedians(as.matrix(VEC))
vecVars<-apply(VEC,1,RowVar)
ratio<-rnaAvg/vecAvg

scatter_dat<-as.data.frame(cbind(rnaAvg,vecAvg))

pdf("median_scatter.pdf")
ggplot(scatter_dat,aes(log2(vecAvg),log2(rnaAvg)))+geom_point()+theme_bw()
dev.off()

# if tag not present in all vector samples, remove before continuing 
vec_na_sum<-rowSums(is.na(VEC))
tags_present<-which(vec_na_sum<dim(VEC)[2])
dat_norm<-dat_norm[tags_present,]


#shuffle 
randRatioShuffle<-function(dat,nshuffle){
  vec<-dat[,grep('vec',names(dat),ignore.case=TRUE)]
  row.names(vec)<-dat$barcodes
  rna<-dat[,grep('vec',names(dat),ignore.case=TRUE,invert=TRUE)]
  row.names(rna)<-rna$barcodes
  rna$barcodes<-NULL
  
  nshuffle<-as.numeric(nshuffle)
  #initialize ratio_output_dataframe 
  ratio_output_dataframe<-data.frame(row.names=row.names(rna))
  rna_out<-data.frame(row.names=row.names(rna))
  vec_out<-data.frame(row.names=row.names(vec))
  rna<-as.matrix(rna)
  vec<-as.matrix(vec)
  for(i in seq(1:nshuffle)){
    #pick 2 rna at random
    rna_curr<-rowMedians(rna[,sample(1:dim(rna)[2],2)])
    #pick 3 vec at random 
    vec_curr<-rowMedians(vec[,sample(1:dim(vec)[2],3)])
    ratio_curr<-rna_curr/vec_curr
    ratio_output_dataframe<-cbind(ratio_output_dataframe,ratio_curr)
    rna_out<-cbind(rna_out,rna_curr)
    vec_out<-cbind(vec_out,vec_curr)
  }
  names(ratio_output_dataframe)<-seq(1:nshuffle)
  ratio_output_dataframe
}

datratioshuffle<-randRatioShuffle(dat_norm_centered,100)
datratioshuffle_short.melt<-melt(datratioshuffle[,1:10])

med_shuffle_dat<-rowMedians(as.matrix(datratioshuffle),na.rm=TRUE)
med_shuffle_dat<-as.data.frame(med_shuffle_dat)
var_shuffle_dat<-apply(datratioshuffle,1,var)
var_shuffle_dat<-as.data.frame(var_shuffle_dat)

row.names(med_shuffle_dat)<-row.names(datratioshuffle)
med_shuffle_dat$barcodes<-row.names(datratioshuffle)
med_shuffle_dat$variance<-var_shuffle_dat$var_shuffle_dat

shuffle.annot<-merge(annotation,med_shuffle_dat,by.x="barcode",by.y="barcodes")
shuffle.annot$type<-"signal"
shuffle.annot$type[grep("promoter",shuffle.annot$name,ignore.case=TRUE)]<-"promoter"
shuffle.annot$type[grep("scramble",shuffle.annot$name,ignore.case=TRUE)]<-"scramble"
locus_shuffle_annot_save<-shuffle.annot
median_ratio<-median(shuffle.annot$med_shuffle_dat,na.rm=TRUE)


#still need to do this?? 
#shuffle.annot$centered_shuffle_ratio<-shuffle.annot$med_shuffle_dat/median_ratio
```

```{r controls}
CMV<-shuffle.annot[grep("CMV",shuffle.annot$name),]
#ggplot(CMV,aes(name,med_shuffle_dat))+geom_boxplot(notch=TRUE)+theme_bw()
#pdf("cmv_controls_all.pdf")
ggplot(CMV,aes(name,log2(med_shuffle_dat)))+geom_boxplot(notch=TRUE)+theme_bw()
#dev.off()
#ggplot(CMV,aes(name,centered_shuffle_ratio))+geom_boxplot(notch=TRUE)+theme_bw()
#ggplot(CMV,aes(name,log2(centered_shuffle_ratio)))+geom_boxplot(notch=TRUE)+theme_bw()

```

Center data
```{r centerdata}
#shuffle.annot$med_shuffle_dat<-shuffle.annot$centered_shuffle_ratio
```


```{r binned_locus_boxplot}
# boxplots of signal across locus, binned 
locus_only_shuffle.annot<-shuffle.annot
locus_only_shuffle.annot<-locus_only_shuffle.annot[grep("promoter",locus_only_shuffle.annot$name,invert=TRUE),]
locus_only_shuffle.annot<-locus_only_shuffle.annot[grep("scramble",locus_only_shuffle.annot$name,invert=TRUE),]

#sort on localOrder 
locus_only_shuffle.annot<-locus_only_shuffle.annot[order(locus_only_shuffle.annot$LocalOrder),]
#partition into 1,100 evenly 
locus_only_shuffle.annot$binnedOrder<-as.factor(cut(locus_only_shuffle.annot$LocalOrder,100,labels=FALSE))
names(locus_only_shuffle.annot)[grep("med_shuffle_dat",names(locus_only_shuffle.annot))]<-"ratio"
```


```{r dotplots}
locus_dat<-locus_only_shuffle.annot
loc<-str_split_fixed(locus_dat$localCoordinates,":",2)
startandend<-str_split_fixed(loc[,2],"-",2)
start<-as.numeric(startandend[,1])
locus_dat$start<-start
locus_dat<-locus_dat[grep("promoter|scrambled",locus_dat$name,ignore.case=TRUE,invert=TRUE),]
unique(locus_dat$name)
```


# sliding window plot

```{r functions_for_sliding_window_plots, echo=FALSE, results='hide'}
library(plyr)
library(ggplot2)
library(stringr)
library(reshape2)
library(matrixStats)
library(gridExtra)
#library(zoo)
library(grid)
library(reshape2)
library(gridBase)

restructure_to_bases_mm10<-function(dat_elements_annot,strand){
  #prepare coordinate data
  locus<-str_split_fixed(dat_elements_annot$localCoordinates,":",2)
  startandend<-str_split_fixed(locus[,2],"-",2)
  chr<-locus[,1]
  start<-startandend[,1]
  end<-startandend[,2]
  dat_elements_annot$chr<-chr
  dat_elements_annot$start<-as.numeric(start)
  dat_elements_annot$end<-as.numeric(end)
  dat_elements_annot<-dat_elements_annot[which(!is.na(dat_elements_annot$start)),] #elements with no genomic coordinates can be removed here. 
  
  #BASES -- melt so that each base (of each oligo) has its own row. --- 5 ratio measurements per base! 
  basesGenomic<-str_split(dat_elements_annot$Element,"")
  basesGenomic<-t(as.matrix(as.data.frame(basesGenomic)))
  basesGenomic<-as.data.frame(basesGenomic)
  row.names(basesGenomic)<-dat_elements_annot$barcode
  basesGenomic$barcode<-row.names(basesGenomic)
  basesGenomicExpanded<-merge(dat_elements_annot,basesGenomic,by.x="barcode",by.y="barcode")
  
  Gbases_melted<-melt(basesGenomicExpanded,id.vars=names(basesGenomicExpanded)[1:19])
  
  #ID VARS ARE EVERYTHING THAT ARE NOT V1...V90 (BASES)
  #FIGURE OUT A BETTER WAY TO DO THIS!!! -___-
  #OTHERWISE, ALWAYS CHECK HERE BC THESE HARD CODED NUMBERS CHANGE EVERY TIME THE OLIGOS CHANGE... 
  Gbases_melted<-Gbases_melted[order(Gbases_melted$LocalOrder),]

  #COORDS
  coords<-as.data.frame(dat_elements_annot[,c("start","end")])
  coords$start<-as.numeric(coords$start)
  coords$end<-as.numeric(coords$end)
  coords<-coords[which(!is.na(coords$start)),]
  
  basecoordsGenomic<-t(apply(coords,1,function(x) seq.int(from=x[1],to=x[2])))
  if(strand=="-"){ #count backwards if neg strand #need to make sure coords count DOWN bc on - strand! ###(to associate proper base to proper coord, even tho it will be the complement of the base as viewed in the genome browser!)
      basecoordsGenomic<-t(apply(coords,1,function(x) seq.int(from=x[2],to=x[1])))
  }
  baseCoordsGenomicExpanded<-cbind(dat_elements_annot,basecoordsGenomic)
  Gbasecoords_melted<-melt(baseCoordsGenomicExpanded,id.vars=names(baseCoordsGenomicExpanded)[1:19])
  Gbasecoords_melted<-Gbasecoords_melted[order(Gbasecoords_melted$LocalOrder),]
    #ID VARS ARE EVERYTHING THAT ARE NOT V1...V90 (base coords)
    #FIGURE OUT A BETTER WAY TO DO THIS!!! -___-
    # OTHERWISE, ALWAYS CHECK HERE BC THESE HARD CODED NUMBERS CHANGE EVERY TIME THE OLIGOS CHANGE...
    #ALSO CHECK STRANDING IS RIGHT (Direction of counting...) 
  
  
  Gbases_melted$variable<-NULL
  names(Gbases_melted)[grep('value',names(Gbases_melted))]<-"Base"
  genomic<-cbind(Gbases_melted,Gbasecoords_melted[,"value"])
  names(genomic)[grep('value',names(genomic))]<-"Coord"
  genomic$uniquePosition<-paste(genomic$chr,genomic$Coord,sep=":")
  
  
  # use unique genomic position (created above) to gather the median ratio
  # in this script ratio input is shuffled ratio (taken from 1000 selections of [mean of 2 random RNA samples/mean of 3 random DNA replicates]). Now select median (na.rm=TRUE) as representative of all 5 tags for this element. 
  
  #genomic_medians_element<-aggregate(ratio~Element,data=genomic,median,na.rm=TRUE)
  genomic_medians_element<-aggregate(ratio~Element,data=genomic,median,drop=FALSE)
  names(genomic_medians_element)[grep("ratio",names(genomic_medians_element))]<-"ElementMedianRatio"
  genomic<-merge(genomic,genomic_medians_element,by.x="Element",by.y="Element",all.x=TRUE)
  
  #for bases that are covered by more than one Element due to tiling, report the mean of the two element-wise medians ** (revisit later)
  #genomic_base_scores<-aggregate(ElementMedianRatio~uniquePosition,data=genomic,mean,na.rm=TRUE)
  #genomic_base_scores<-aggregate(ElementMedianRatio~uniquePosition,data=genomic,mean,drop=FALSE) #dropping some values... 
  
  genomic_base_scores<-cast(genomic,uniquePosition~.,fun=mean)
  names(genomic_base_scores)<-c("uniquePosition","MeanElementMedianRatio")
  
  #reintroduce element,barcode,name,localorder,uniqueposition,base,coord
  newannotation<-genomic[,c("Element","barcode","name","LocalOrder","uniquePosition","Base","Coord")]
  newannotation<-ddply(newannotation,.(uniquePosition),head,n=1)
  new_genomic<-merge(newannotation,genomic_base_scores,by.x="uniquePosition",by.y="uniquePosition")
  
  #organize final data and return 
  new_genomic<-new_genomic[order(new_genomic$Coord),]
  if(strand=="-"){new_genomic<-new_genomic[order(new_genomic$Coord,decreasing=TRUE),]}
  # check that localorder 1 is first
  # check base by base that coord matches base in element (if "-", this will be from end of "element")
  new_genomic
}

# sliding window
# Modified from http://coleoguy.blogspot.com/2014/04/sliding-window-analysis.html
# similar to rollapply but faster
# updated to reflect ElementMedianRatio name change 1-10-17

slideFunct <- function(data, window, step){
  total <- dim(data)[1]
  window<-as.numeric(window)
  spots <- seq(from = 1, to = (total - window + 1), by = step)
  rolledRatio<-vector(length = length(spots))
  uniquePosition<-vector(length = length(spots))
  for(i in 1:length(spots)){
    rolledRatio[i]<-mean(data$MeanElementMedianRatio[spots[i]:(spots[i]+window-1)])
    uniquePosition[i]<-data$uniquePosition[spots[i]] 
  }
  result.df<-as.data.frame(cbind(rolledRatio,uniquePosition))
  colnames(result.df)<-c("rolledRatio","uniquePosition")
  result.df$rolledRatio<-as.numeric(as.character(result.df$rolledRatio))
  return(result.df)
}

#scan statistic 
#input<-d, window, step, permutedmatrix, number of randomization used, dat (with bases for peakseq) 
#output:
#list: d_bins_copy, ggplot_up, peaklist 

statScanSlideWindows<-function(dat,window,step,permutedMatrix,randomize_reps,strand=strand){
  #ensure data is properly ordered
  dat<-dat[order(dat$Coord),]
  if(strand=="-"){dat<-dat[order(dat$Coord,decreasing=TRUE),]}
  
  #assume possible multiple loci input, break into list of loci and do slidefunction on each
  dfList<-split(dat,dat$name)
  dfList<-dfList[lapply(dfList,function(x)dim(x)[1])>0]
  binList<-lapply(dfList,function(x)slideFunct(x,window,step))
  binList<-do.call("rbind",binList)
  
  d_bins<-merge(dat,binList,by.x="uniquePosition",by.y="uniquePosition")
  rm(binList)
  rm(dfList)
  d_bins$chr<-str_split_fixed(d_bins$uniquePosition,":",2)[,1]
  d_bins$end<-d_bins$Coord+window-1 #will need to correct for if this is out of bounds of the locus
  d_bins$score<-0
  d_bins$pval<-0

  #sigUP
  for(i in seq(1:randomize_reps)){
    dperm<-dat;
    dperm$MeanElementMedianRatio<-permutedMatrix[,i]
    dpermList<-split(dperm,dperm$name)
    dpermList<-dpermList[lapply(dpermList,function(x)dim(x)[1])>0]
    dperm_bins<-lapply(dpermList,function(x)slideFunct(x,window,step))
    dperm_bins<-do.call("rbind",dperm_bins)
    d_bins$score[which(dperm_bins$rolledRatio>=d_bins$rolledRatio)]<-(d_bins$score[which(dperm_bins$rolledRatio>=d_bins$rolledRatio)]+1)
    d_bins$pval<-d_bins$score/i
  }
  d_bins$sig<-"no"
  
  #benjamini hochberg correction!
  # order pvalues from smallest to largest
  d_bins<-d_bins[order(d_bins$pval),]
  #assign ranks 
  d_bins$pvalrank<-seq(from=1,to=length(d_bins$pval))
  #calc q vals
  m=length(d_bins$pval)
  FDR=0.01 #1% FDR
  d_bins$BH_qval<-(d_bins$pvalrank/m)*FDR
  #find largest p<q
  plessthanq<-d_bins$BH_qval-d_bins$pval #highest ranking positive val 
  
  #if any pos values in plessthanq:
  if(any(plessthanq>0)){
      cutoff_index<-max(which(plessthanq>0))
      d_bins$sig[1:cutoff_index]<-"yes"    #everything above this is significant!
      up_signal<-d_bins[which(d_bins$sig=="yes"),]
  }else{#else: nothing significant.. 
    up_signal<-NA
    }
  #revert to coordinate-order 
  d_bins<-d_bins[order(d_bins$Coord),]
  if(strand=="-"){d_bins<-d_bins[order(d_bins$Coord,decreasing=TRUE),]}
  
  #if coord distance < window size from previous coord distance, label as same peak. otherwise increment by 1 
  
  d_bins$upPeaks<-factor(paste(d_bins$name,d_bins$sig,sep="_"))
  peaks_up<-split(d_bins,d_bins$upPeaks)
  peaks_up<-peaks_up[grep("*yes",names(peaks_up))]
  
  
  peaks_up<-lapply(peaks_up,function(x){
    length<-dim(x)[1]
    x$Coord<-as.numeric(as.character(x$Coord))
    peaknum<-1
    peakList<-rep(1,length)
    x<-x[order(x$name,x$Coord),]
    
    if(length<=2){
        if(length==2){
          if(abs(x$Coord[2]-x$Coord[1])>window){peakList<-c(1,2)} 
          #if(abs(x$Coord[2]-x$Coord[1])>2*step){peakList<-c(1,2)}
        }
        x$peakNums<-peakList
        x
    }else{
      for(i in (2:length)){
        if(abs(x$Coord[i]-x$Coord[i-1])>window){peaknum<-peaknum+1}
        peakList[i]<-peaknum
      }
      x$peakNums<-as.numeric(as.character(peakList))
      x
    }
  })
  up<-do.call("rbind",peaks_up)
  
  #if up peaks exist, then merge
  test<-ifelse(length(up)>=1,"TRUE","FALSE")
  if(test=="TRUE"){
    d_bins<-merge(d_bins,up[,c("uniquePosition","peakNums")],all.x=TRUE)
  }else{
    d_bins$peakNums<-0
  }
  rm(up)
    
  A<-ggplot(d_bins,aes(xmin=Coord,xmax=Coord+500,ymin=0,ymax=log2(rolledRatio),fill=sig))+geom_rect(aes(width=500))+facet_grid(.~name,scales="free")+theme_bw()
  #A<-ggplot(d_bins,aes(Coord,log2(rolledRatio),colour=sig))+geom_point()+theme_bw() #+facet_grid(.~name,scales="free")

    
  d_binsup<-d_bins[order(d_bins$Coord),]
  
  d_binsup<-d_binsup[which(d_binsup$peakNums>0),]
  peaks<-split(d_binsup,d_binsup$peakNums)
  input_bed<-vector()
  input_bed<-lapply(peaks,function(x){
    start<-x[1,7]
    end<-x[dim(x)[[1]],11]
    chr<-ifelse(is.na(x[1,10]),"NotGenomic",x[1,10])
    input_bed<-rbind(input_bed,c(chr,start,end)) #chr, start, end of peak 
    
    #start<-x[1,7]
    #end<-x[dim(x)[[1]],12]
    #chr<-ifelse(is.na(x[1,11]),"NotGenomic",x[1,11])
    #input_bed<-rbind(input_bed,c(chr,start,end)) #chr, start, end of peak 

  }
  )
  test<-ifelse(length(input_bed)>=1,"TRUE","FALSE")
  if(test=="TRUE"){
    input_bed<-as.data.frame(do.call("rbind",input_bed))
    names(input_bed)<-c("chr","start","end")
    input_bed<-as.data.frame(input_bed)
  }else{
      input_bed<-"NO PEAKS"
  }
  return(list(d_bins,A,input_bed))
  }

```




```{r process_sliding_window_data}
#really just want the MEDIANS of the individual tag ratios--dont want to do improper comparisons 
locus_dat$variance<-NULL #when originally wrote this did not have var column so woul dhave to re-write to accomodate this column otherwise this breaks -_- 
locus_dat$centered_shuffle_ratio<-NULL
names(locus_dat)[grep("med_shuffle_dat",names(locus_dat))]<-"ratio"

locus_dat<-locus_dat[order(locus_dat$name,locus_dat$LocalOrder),]
locus_dat_save<-locus_dat

locus_dat_list<-split(locus_dat,locus_dat$name)


#build dpermuted matrix by permuting sample labels NOT base position labels... 
randLabelRatioShuffle<-function(dat_norm_input,shuffle_input,nshuffle,strand,d){
  dat_norm_input<-dat_norm_input[which(dat_norm_input$barcodes %in% shuffle_input$barcode),]
  row.names(dat_norm_input)<-dat_norm_input$barcodes
  dat_norm_input$barcodes<-NULL
  
  nshuffle<-as.numeric(nshuffle)
  
  #initialize dpermuted_matrix 
  dpermuted_matrix<-data.frame(row.names=row.names(d))
  
  for(i in seq(1:nshuffle)){
    #pick 2 rna at random
    rna_curr<-rowMeans(dat_norm_input[,round(runif(2,min=1,max=dim(dat_norm_input)[2]))])
    #pick 3 vec at random 
    vec_curr<-rowMeans(dat_norm_input[,round(runif(3,min=1,max=dim(dat_norm_input)[2]))])
    
    ratio_curr<-rna_curr/vec_curr

    shuffle_input$ratio<-ratio_curr
    curr_base_dat<-restructure_to_bases_mm10(shuffle_input,strand)
    dpermuted_matrix<-cbind(dpermuted_matrix,curr_base_dat$MeanElementMedianRatio)
    
  }
  names(dpermuted_matrix)<-seq(1:nshuffle)
  dpermuted_matrix
}



set.seed(1)
randomize_reps<-1000

#prod<-lapply(locus_dat_list,function(x){
#  locus_dat_curr<-x
#  strand<-locus_dat_curr$strand[[1]]
#  base_dat<-restructure_to_bases_mm10(locus_dat_curr,strand)
#  base_dat$strand<-rep(strand,dim(base_dat)[1])
#  dpermutedMat_curr<-randLabelRatioShuffle(dat_norm,locus_dat_curr,randomize_reps,strand,base_dat)
#  list(base_dat,dpermutedMat_curr)
#})
#all_base_dats_list<-sapply(prod,function(x) x[1])
#all_loci_d2<-do.call("rbind",all_base_dats_list)
#all_dperm_list<-sapply(prod,function(x) x[2])
#all_loci_permutedMat<-do.call("rbind",all_dperm_list)


#save(prod,file="/n/rinn_data2/users/agroff/oligogames/HiSeq/2017_lincmRNA_C2C12/analysis/Summary_MPRA_analysis_all_loci_basedatsanddperm.RData")






#make prod list by reading in individually saved files 
analysis_dir<-"/Volumes/valor2/users/agroff/oligogames/HiSeq/2017_lincmRNA_C2C12/analysis/"
base_dats <- list.files(analysis_dir,pattern="*base_dat.RData",full.names=TRUE)
permMats <- list.files(analysis_dir,pattern="*Matrix.RData",full.names=TRUE)
filenames_permats<- list.files(analysis_dir,pattern="*Matrix.RData")
filenames_permats<-unlist(filenames_permats)
filenames_permats_exp<-as.data.frame(str_split_fixed(filenames_permats,"_",2))
filenames_permats<-unlist(filenames_permats_exp[,1])
permMatLists<-lapply(permMats,function(x) mget(load(x)))
names(permMatLists)<-unlist(filenames_permats)


filenames_base<- list.files(analysis_dir,pattern="*base_dat.RData")
filenames_base<-unlist(filenames_base)
filenames_base_exp<-as.data.frame(str_split_fixed(filenames_base,"_",2))
filenames_base<-unlist(filenames_base_exp[,1])
basedatList<-lapply(base_dats,function(x) mget(load(x)))
names(basedatList)<-filenames_base


myColors<-c(no="black",yes="red")

numOfFiles<-length(filenames_permats)
output_dbins<-vector("list",numOfFiles)
output_figs<-vector("list",numOfFiles)
output_peaks<-vector("list",numOfFiles)
output_peak_seqs<-vector("list",numOfFiles)

for(i in seq(1,numOfFiles)){
#for(i in seq(1,2)){
  curr_file_name<-as.character(filenames_base[i])
  d<-basedatList[curr_file_name][[1]][[1]]
  strand<-d$strand[[1]];
  d$strand<-NULL;
  dpermutedMatrix<-permMatLists[[curr_file_name]][[1]];
  print(curr_file_name)
  out<-statScanSlideWindows(d,500,50,dpermutedMatrix,randomize_reps,strand)
  output_dbins[i]<-out[1]
  output_figs[i]<-out[2]
  peaks_curr<-out[[3]]
  peaks_curr$name<-curr_file_name
  output_peaks[i]<-list(peaks_curr) #if this is broken its in statscanwindow--somehow shifted the columns.. 
  if(length(dim(peaks_curr)[1])>0){
    peak_seqs_curr<-vector("list",dim(peaks_curr)[1])
    for(j in seq(1,dim(peaks_curr)[1])){
        #if strand is negative, do I need to invert? 
        curr_peak_start_coord<-as.numeric(as.character(peaks_curr$start[j]))
        curr_peak_end_coord<-as.numeric(as.character(peaks_curr$end[j]))
    
        #need to show missing bases in peaks... 
        #make a df with abs(coordstart-coordend) rows and populate w N
        peak_seq_df<-data.frame(seq.int(from=curr_peak_start_coord,curr_peak_end_coord),"N")
        colnames(peak_seq_df)<-c("coords","bases")
        #grab bases for occuring coords 
        peak_seq_df2<-merge(d,peak_seq_df,by.x="Coord",by.y="coords",all.y=TRUE)
        peak_seq_df2$Base<-as.character(peak_seq_df2$Base)
        peak_seq_df2<-peak_seq_df2[,c("Coord","Base")]
        peak_seq_df2[is.na(peak_seq_df2)]<-"N"
  
        peak_seq_curr<-paste(peak_seq_df2$Base,"",collapse="")
        peak_seq_curr<-gsub(" ","",peak_seq_curr)
        if(strand=="-"){
          peak_vec<-peak_seq_df2$Base
          rev_peak<-rev(peak_vec)
          peak_seq_curr<-paste(rev_peak,collapse="")
        }
        curr_peak_seq<-list(peak_seq_curr)
        names(curr_peak_seq)<-paste(curr_file_name,"peak",j,sep="_")
        peak_seqs_curr[j]<-list(curr_peak_seq)
    }
      
    output_peak_seqs[i]<-list(peak_seqs_curr)
  }else{
    output_peak_seqs[i]<-list(c("nopeaks"))
  }
  names(output_peak_seqs)[i]<-curr_file_name
}


#DOTS:
#output_figs[[10]]+ylim(-2,4)+scale_colour_manual(values=myColors)
#bars:
#output_figs[[10]]+ylim(-2,4)+scale_fill_manual(values=myColors)

update_figs<-lapply(output_figs,function(x) x+ylim(-2,7)+scale_fill_manual(values=myColors))
library(gridExtra)
pdf("011418_all_locus_plots.pdf",height=11,width=24)
do.call("grid.arrange",update_figs) 
dev.off()

#pdf(paste(sample,linc,"peaks.pdf",sep="_"))
#outputList[[2]]+ggtitle(paste(linc, "window 500, step 50"))+geom_vline(xintercept=TSS_coords)+ylim(-2,4)+scale_fill_manual(values=myColors)
#dev.off()

#d_bins<-outputList[[1]]
#sigregions<-d_bins[which(d_bins$sig=="yes"),]
#meansig<-mean(sigregions$rolledRatio,na.rm=TRUE)
#maxsig<-max(sigregions$rolledRatio,na.rm=TRUE)

#`r locus` mean sig value: `r meansig`
#`r locus` max sig value: `r maxsig`


#concatenate all outputLists[[3]] to get bed of mm10 peaks -- get DNA from ucsc and run fimo together 
output_peaks[5]<-NULL #foxf1 no peaks 
peakslist<-do.call("rbind",output_peaks)
peakslist$start<-as.numeric(as.character(peakslist$start))
peakslist$end<-as.numeric(as.character(peakslist$end))
peakslist$length<-abs(peakslist$start-peakslist$end)
peakslist$type<-"mRNA"
peakslist$type[which(peakslist$name %in% c("Peril","Crnde","Fendrr","Lincenc1","Lincp21","Tug1"))]<-"lincRNA"
num<-split(peakslist,peakslist$type)
peak_nums<-lapply(num,dim)
meanLengths<-aggregate(peakslist,by=list(peakslist$type),mean)

peaknums<-as.data.frame(peak_nums)
peaknums2<-peaknums[1,]
peaknums<-peaknums2
peaknums.melt<-melt(peaknums)
pdf("peaknums.pdf")
ggplot(peaknums.melt,aes(variable,value,fill=variable))+geom_bar(stat="identity")+theme_bw()+scale_fill_manual(values=c("lincRNA"="red","mRNA"="black"))
dev.off()

pdf("peaklength.pdf")
ggplot(peakslist,aes(type,length,colour=type))+geom_boxplot()+theme_bw()+scale_colour_manual(values=c("lincRNA"="red","mRNA"="black"))
dev.off()
```




# MPRA Summary trend analysis- CDFs 
Setup/load data 

```{r setup, include=FALSE}
.libPaths( c("/n/home06/agroff/R/x86_64-unknown-linux-gnu-library/3.3",.libPaths()))
library(knitr)
library(rmarkdown)
library(ggplot2)
library(reshape)
library(plyr)
library(Biobase)
library(stringr)
library(reshape2)

opts_chunk$set(echo=FALSE, message=FALSE, warning=FALSE,fig.height=8, fig.width=8, dev=c('png', 'pdf'))

data_counts_orig<-read.table("/Volumes/valor2/users/agroff/oligogames/HiSeq/2017_lincmRNA_C2C12/analysis/C2C12_lincmRNA_combined_Jan2017_hiseq.tab",header=TRUE,stringsAsFactors = FALSE)
annotation<-read.table("/Volumes/valor2/users/agroff/oligogames/HiSeq/2017_lincmRNA_C2C12/analysis/lincmRNA_pool_annotation_updated_promotersinloci.txt",header=TRUE,stringsAsFactors = FALSE)
locus_TSS<-read.table("/Volumes/valor2/users/agroff/oligogames/HiSeq/2017_lincmRNA_C2C12/analysis/lincmRNA_TSSs.txt",header=TRUE,stringsAsFactors = FALSE,sep="\t")

threshold<-10 

QC.dat.annot<-merge(annotation,data_counts_orig,by.x="barcode",by.y="barcodes")

normalize_andcenter<-function(dat,threshold){
  dat2<-as.matrix(dat[,2:dim(dat)[2]])
  dat2[dat2<threshold]<-NA
  repsums<-colSums(dat2,na.rm=TRUE)
  dat_norm<-t(t(dat2)/repsums)
  medians<-apply(dat_norm,2,median,na.rm=TRUE)
  dat_norm2<-t(t(dat_norm)/medians)
  dat<-cbind(dat["barcodes"],dat_norm2)
  #namesfordat<-c("barcodes",names(dat_norm))
  as.data.frame(dat)
  #names(dat)<-namesfordat
  #dat
}
#dat_norm<-normalize(data_counts_touse,threshold)
dat_norm<-normalize_andcenter(data_counts_orig,threshold)
set.seed(1)

```

Get ratios

```{r getratios}
dat_norm2<-dat_norm[,1:11]
colnames(dat_norm2)<-c("barcodes","C2C12_1","C2C12_2","C2C12_3","C2C12_4","C2C12_5","VEC_1","VEC_2","VEC_3","VEC_4","VEC_5")
dat_norm<-dat_norm2;rm(dat_norm2)
dat_norm.melt<-melt(dat_norm)

info<-str_split_fixed(dat_norm.melt$variable,"_",2)
dat_norm.melt$sampletype<-info[,1]
dat_norm.melt$replicate<-info[,2]


RowVar <- function(x) {
  ans<-sum((x - mean(x,na.rm=TRUE)^2),na.rm=TRUE)/(length(x) - 1)
  ans
}

#change name to eventually make function 
names(dat_norm)<-gsub("C2C12","RNA",names(dat_norm))
names(dat_norm)<-gsub("lincmRNAvec","VEC",names(dat_norm))

RNA<-dat_norm[,grep("RNA",names(dat_norm))]
row.names(RNA)<-dat_norm$barcodes
VEC<-dat_norm[,grep("VEC",names(dat_norm))]
row.names(VEC)<-dat_norm$barcodes


# if tag not present in all vector samples, remove before continuing 
vec_na_sum<-rowSums(is.na(VEC))
tags_present<-which(vec_na_sum<dim(VEC)[2])
dat_norm<-dat_norm[tags_present,]


#shuffle 
randRatioShuffle<-function(dat,nshuffle){
  vec<-dat[,grep('vec',names(dat),ignore.case=TRUE)]
  row.names(vec)<-dat$barcodes
  rna<-dat[,grep('vec',names(dat),ignore.case=TRUE,invert=TRUE)]
  row.names(rna)<-rna$barcodes
  rna$barcodes<-NULL
  
  nshuffle<-as.numeric(nshuffle)
  #initialize ratio_output_dataframe 
  ratio_output_dataframe<-data.frame(row.names=row.names(rna))
  rna_out<-data.frame(row.names=row.names(rna))
  vec_out<-data.frame(row.names=row.names(vec))
  rna<-as.matrix(rna)
  vec<-as.matrix(vec)
  for(i in seq(1:nshuffle)){
    #pick 2 rna at random
    rna_curr<-rowMedians(rna[,sample(1:dim(rna)[2],2)])
    #pick 3 vec at random 
    vec_curr<-rowMedians(vec[,sample(1:dim(vec)[2],3)])
    ratio_curr<-rna_curr/vec_curr
    ratio_output_dataframe<-cbind(ratio_output_dataframe,ratio_curr)
    rna_out<-cbind(rna_out,rna_curr)
    vec_out<-cbind(vec_out,vec_curr)
  }
  names(ratio_output_dataframe)<-seq(1:nshuffle)
  #rna_sd<-apply(rna_out,1,sd)
  #vec_sd<-apply(rna_out,1,sd)
  #rna_mean<-rowMeans(rna_out)
  #vec_mean<-rowMeans(vec_out)
  #cov_rna_vec<-diag(cov(t(rna_out),t(vec_out))) #row-wise covariance of rna,vec
  #var_ratio<-(rna_mean^2/vec_mean^2)*((rna_sd^2/rna_mean^2)-(2*cov_rna_vec/(rna_mean*vec_mean))+(vec_sd^2/vec_mean^2))
  #var_ratio<-var(ratio_output_dataframe)
  #list(ratio_output_dataframe,var_ratio)
  as.data.frame(ratio_output_dataframe)
}
datratioshuffle<-randRatioShuffle(dat_norm,1000)
med_shuffle_dat<-rowMedians(as.matrix(datratioshuffle),na.rm=TRUE)
med_shuffle_dat<-as.data.frame(med_shuffle_dat)
var_shuffle_dat<-apply(datratioshuffle,1,var)
var_shuffle_dat<-as.data.frame(var_shuffle_dat)
row.names(med_shuffle_dat)<-row.names(datratioshuffle)
med_shuffle_dat$barcodes<-row.names(datratioshuffle)
med_shuffle_dat$var<-var_shuffle_dat$var_shuffle_dat
shuffle.annot<-merge(annotation,med_shuffle_dat,by.x="barcode",by.y="barcodes")
shuffle.annot$type<-"signal"
shuffle.annot$type[grep("promoter",shuffle.annot$name,ignore.case=TRUE)]<-"promoter"
shuffle.annot$type[grep("scramble",shuffle.annot$name,ignore.case=TRUE)]<-"scramble"
locus_shuffle_annot_save<-shuffle.annot
```

TSS labels 

```{r TSS_labels}
locus_TSS<-read.table("/Volumes/valor2/users/agroff/oligogames/HiSeq/2017_lincmRNA_C2C12/analysis/lincmRNA_TSSs.txt",header=TRUE,stringsAsFactors = FALSE,sep="\t")
locus_TSS<-locus_TSS[order(locus_TSS$chr,locus_TSS$TSS_coord),]
locus_TSS$name<-gsub(";","",locus_TSS$name)

#oligos_dat2<-shuffle.annot[which(shuffle.annot$med_shuffle_dat>2),]
oligos_dat2<-shuffle.annot
oligos_dat2<-oligos_dat2[grep("promoter|scramble|Super",oligos_dat2$name,invert=TRUE),]
oligos_dat2$genename<-str_split_fixed(oligos_dat2$name,"_",3)[,1]
oligos_dat2$genename<-gsub("Enc1","Enc1_mm10_refGene",oligos_dat2$genename)

#label multiple TSSs 
locus_TSS$uniqueTSSID<-seq(1,dim(locus_TSS)[1])

#merge tss by genename in oligos_dat
oligodat2<-merge(oligos_dat2,locus_TSS,by.x="genename",by.y="name",all=TRUE)
coords<-str_split_fixed(oligodat2$localCoordinates,":",2)[,2]
oligodat2$localstart<-str_split_fixed(coords,"-",2)[,1]

#handle strandedness 
strands<-split(oligodat2,oligodat2$strand.x)

negstrandoligos<-strands[[1]]
negstrandoligos$distance<-as.numeric(as.character(negstrandoligos$TSS_coord))-as.numeric(as.character(negstrandoligos$localstart))
posstrandoligos<-strands[[2]]
posstrandoligos$distance<-as.numeric(as.character(posstrandoligos$localstart))-as.numeric(as.character(posstrandoligos$TSS_coord))

oligodat3<-rbind(negstrandoligos,posstrandoligos)
oligodat3$location<-"far"
oligodat3$location[which(oligodat3$distance<0 & oligodat3$distance>(-1000))]<-"near"
#oligodat3$location[which(oligodat3$distance<0 & oligodat3$distance>(-500))]<-"near"
oligodat3<-oligodat3[order(oligodat3$location,oligodat3$distance,decreasing=TRUE),]
oligodat3<-ddply(oligodat3,.(barcode),head,1)
oligodat3<-oligodat3[order(oligodat3$distance),]

#combine, sort so all nears appear first, and then again keep only one observation per element 
#(ie either near any tss, or not)

oligodat3$genetype<-"NA"
oligodat3$genetype[grep("Lincenc1|Lincp21|Fendrr|Tug1|Crnde|Peril",oligodat3$genename)]<-"lincRNA"
oligodat3$genetype[grep("Morc2a|Cdkn1a|Foxf1|Sox2|Enc1_mm10|Irx5",oligodat3$genename)]<-"mRNA"

#all oligos within 500bp of a TSS vs 1000 (either side) or more 
oligodat3$disbins<-"genebody"
oligodat3$disbins[which(abs(oligodat3$distance)<=1000)]<-"promoter"

#oligodat3$expressed<-"Expressed"
#oligodat3$expressed[grep("Lincp21|Lincenc1|Peril|Fendrr|Sox2|Irx5|Foxf1",oligodat3$genename)]<-"NotExpressed"

```

CDFs

```{r cdfs}
#cdf signal linc or mRNA promoter or not 
#pdf("lincvmRNA_genebodyvpromoter_cdf.pdf")
#ggplot(oligodat3, aes(log2(med_shuffle_dat),colour=disbins)) + stat_ecdf(geom = "step") + facet_wrap(~genetype)+theme_bw()+ggtitle("lincrna v mrna")+scale_colour_manual(values=c("promoter"="black","genebody"="grey"))
#dev.off()

#pdf("lincvmRNA_genebodyvpromoter_zoom_cdf.pdf")
#ggplot(oligodat3, aes(log2(med_shuffle_dat),colour=disbins)) + stat_ecdf(geom = "step") + facet_wrap(~genetype)+theme_bw()+ggtitle("lincrna v mrna")+scale_colour_manual(values=c("promoter"="black","genebody"="grey"))+coord_cartesian(xlim = c(-1,1))
#dev.off()
#ggplot(oligodat3, aes(disbins,log2(med_shuffle_dat),colour=disbins)) + geom_violin()+ facet_wrap(~genetype)+theme_bw()+ggtitle("lincrna v mrna")+scale_colour_manual(values=c("promoter"="black","genebody"="grey"))#+coord_cartesian(ylim = c(-1,1))

pdf("lincvmRNA_genebodyvpromoter_boxplots.pdf")
ggplot(oligodat3, aes(disbins,log2(med_shuffle_dat),colour=disbins)) + geom_boxplot(notch=TRUE)+ facet_wrap(~genetype)+theme_bw()+ggtitle("lincrna v mrna")+scale_colour_manual(values=c("promoter"="black","genebody"="grey"))#+coord_cartesian(ylim = c(-1,1))
dev.off()





#ks.test promoter v genebody mRNA
mRNAs<-oligodat3[which(oligodat3$genetype=="mRNA"),]
#ks.test(mRNAs[which(mRNAs$disbins=="promoter"),"med_shuffle_dat"],mRNAs[which(mRNAs$disbins=="genebody"),"med_shuffle_dat"])
#reject! these distributions are not equal 

#ks.test promoter v genebody lincRNA
lincRNAs<-oligodat3[which(oligodat3$genetype=="lincRNA"),]
#ks.test(lincRNAs[which(lincRNAs$disbins=="promoter"),"med_shuffle_dat"],lincRNAs[which(lincRNAs$disbins=="genebody"),"med_shuffle_dat"])

```


```{r sessioninfo}
sessionInfo()
```



Fimo Plot each tissue

```{r fimoplot_c2c12_example}

nameconversiontable<-read.table("Core_Jaspar_nameConversionTable.txt")
names(nameconversiontable)<-c("motifname","name")
nameconversiontable$motifname<-gsub(">","",nameconversiontable$motifname)


C2C12fimo<-read.table("C2C12/fimo.txt",header=FALSE)
names(C2C12fimo)<-c("motif","seq","start","end","strand","NA","pvalue","qvalue","matchedseq")
C2C12fimo$sample<-"C2C12"
C2C12fimo$uniqueID<-paste(C2C12fimo$motif,C2C12fimo$start,sep="_")

brainfimo<-read.table("BRAIN/fimo.txt",header=FALSE)
names(brainfimo)<-c("motif","seq","start","end","strand","NA","pvalue","qvalue","matchedseq")
brainfimo$sample<-"WholeBrain"
brainfimo$uniqueID<-paste(brainfimo$motif,brainfimo$start,sep="_")


GEOBfimo<-read.table("GEOB/fimo.txt",header=FALSE)
names(GEOBfimo)<-c("motif","seq","start","end","strand","NA","pvalue","qvalue","matchedseq")
GEOBfimo$sample<-"GEOB"
GEOBfimo$uniqueID<-paste(GEOBfimo$motif,GEOBfimo$start,sep="_")


kidneyfimo<-read.table("Kidney/fimo.txt",header=FALSE)
names(kidneyfimo)<-c("motif","seq","start","end","strand","NA","pvalue","qvalue","matchedseq")
kidneyfimo$sample<-"Kidney"
kidneyfimo$uniqueID<-paste(kidneyfimo$motif,kidneyfimo$start,sep="_")



mESCfimo<-read.table("mESC/fimo.txt",header=FALSE)
names(mESCfimo)<-c("motif","seq","start","end","strand","NA","pvalue","qvalue","matchedseq")
mESCfimo$sample<-"mESC"
mESCfimo$uniqueID<-paste(mESCfimo$motif,mESCfimo$start,sep="_")


all_samples<-rbind(C2C12fimo,brainfimo,GEOBfimo,kidneyfimo,mESCfimo)

fimores_annot<-merge(all_samples,nameconversiontable,by.x="motif",by.y="motifname")
fimores_annot$motifwidth<-with(fimores_annot,end-start)

ggplot(fimores_annot,aes(start,name,fill=qvalue))+geom_tile(width=fimores_annot$motifwidth)+theme_bw()+facet_wrap(~sample)


pdf("motifs_peril_peak3_eachsampletype_q0-05.pdf")
ggplot(subset(fimores_annot,qvalue<0.05),aes(start,name,fill=qvalue,width=motifwidth))+geom_tile()+theme_bw()+facet_wrap(~sample)
dev.off()


#############################

#which motifs occur in both C2C12 and Kidney? 
C2C12_and_kidney<-C2C12fimo[which(C2C12fimo$uniqueID %in% kidneyfimo$uniqueID),]
C2C12_and_kidney_annot<-merge(C2C12_and_kidney,nameconversiontable,by.x="motif",by.y="motifname")
C2C12_and_kidney_annot$motifwidth<-with(C2C12_and_kidney_annot,end-start)

#kidney_and_C2C12<-kidneyfimo[which(kidneyfimo$uniqueID %in% C2C12fimo$uniqueID),]#sanitycheck
subset(C2C12_and_kidney_annot,qvalue<0.05)
ggplot(C2C12_and_kidney_annot,aes(start,name,fill=qvalue))+geom_tile(width=C2C12_and_kidney_annot$motifwidth)+theme_bw()
ggplot(subset(C2C12_and_kidney_annot,qvalue<0.05),aes(start,name,fill=qvalue,width=motifwidth))+geom_tile()+theme_bw()


###########################################
# which motifs occur in all samples? 
in_all_samples<-intersect(C2C12_and_kidney$uniqueID,intersect(mESCfimo$uniqueID,intersect(GEOBfimo$uniqueID,brainfimo$uniqueID)))
common_motifs<-C2C12_and_kidney_annot[which(C2C12_and_kidney_annot$uniqueID %in% in_all_samples),]
common_motifs$motifwidth<-with(common_motifs,end-start)

subset(common_motifs,qvalue<0.05)


ggplot(common_motifs,aes(start,name,fill=qvalue))+geom_tile(width=common_motifs$motifwidth)+theme_bw()

pdf("Common_motifs_PerilPeak3_q0-05.pdf")
ggplot(subset(common_motifs,qvalue<0.05),aes(start,name,fill=qvalue,width=motifwidth))+geom_tile()+theme_bw()
dev.off()
```