mapapaQTL.Rmd

---
title: "apaQTLs"
author: "Briana Mittleman"
date: "4/18/2019"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```


```{r}
library(tidyverse)
library(reshape2)
library(workflowr)
library(cowplot)
```

In this analysis I will call apaQTls in both fractions. I will start with the phenotype files and normalized the counts using the leafcutter package in order to run the fastq QTL mapper.  


##Prepare phenotypes for QTL- phenotype dir  
It is best to run this analysis in the data/phenotype_5perc directory. I have copied the leafcutter prepare_phenotype_table.py to the code directroy to use here.  


```{bash,eval=F}
#!/bin/bash
module load python

gzip APApeak_Phenotype_GeneLocAnno.Total.5perc.fc
gzip APApeak_Phenotype_GeneLocAnno.Nuclear.5perc.fc

python ../../code/prepare_phenotype_table.py APApeak_Phenotype_GeneLocAnno.Total.5perc.fc.gz
python ../../code/prepare_phenotype_table.py APApeak_Phenotype_GeneLocAnno.Nuclear.5perc.fc.gz

```


This will output bash scripts to run. 

```{bash,eval=F}
module load Anaconda3
source activate three-prime-env

sh APApeak_Phenotype_GeneLocAnno.Nuclear.5perc.fc.gz_prepare.sh
sh APApeak_Phenotype_GeneLocAnno.Total.5perc.fc.gz_prepare.sh

```


Subset the PCs to use the first 2 in the qtl calling:  

```{bash,eval=F}
module load Anaconda3
source activate three-prime-env

head -n 3 APApeak_Phenotype_GeneLocAnno.Nuclear.5perc.fc.gz.PCs > APApeak_Phenotype_GeneLocAnno.Nuclear.5perc.fc.gz.2PCs
head -n 3 APApeak_Phenotype_GeneLocAnno.Total.5perc.fc.gz.PCs > APApeak_Phenotype_GeneLocAnno.Total.5perc.fc.gz.2PCs

```

##Call QTLs- code dir  

Next I will need to make a sample list. From the code directory: 

```{bash,eval=F}
python makeSampleList.py
```

*remove 19092 and 19193*
 
 
Prepare directroy 
```{bash,eval=F}
mkdir ../data/apaQTLNominal
mkdir ../data/apaQTLPermuted
```


Run the code to call QTLs within 1mb of each PAS peak. I run both a nominal pass and a permuted pas. The permulted pas chosses the best snp for each peak gene pair. 


```{bash,eval=F}
sbatch apaQTL_Nominal.sh
sbatch apaQTL_permuted.sh
```


Concatinate all of the results in the permuted set. I do this so I can account for multiple testing with the benjamini hochberg test.  

Concatinate results in permuted directory:  
```{bash,eval=F}
cat APApeak_Phenotype_GeneLocAnno.Total.5perc.fc.gz.qqnorm_chr* > APApeak_Phenotype_GeneLocAnno.Total_permRes.txt

cat APApeak_Phenotype_GeneLocAnno.Nuclear.5perc.fc.gz.qqnorm_chr* > APApeak_Phenotype_GeneLocAnno.Nuclear_permRes.txt
 
```


Run correction script  
```{bash,eval=F}
Rscripts apaQTLCorrectPvalMakeQQ.R  
```


##Evaluation results   

```{r}
totRes=read.table("../data/apaQTLPermuted/APApeak_Phenotype_GeneLocAnno.Total_permResBH.txt", stringsAsFactors = F, header = T) %>% separate(pid, into=c("Chr", "Start", "End", "PeakID"), sep=":") %>% separate(PeakID, into=c("Gene", "Loc", "Strand","Peak"), sep="_")
```


Total Apa QTLs   
```{r}
TotQTLs= totRes %>% filter(-log10(bh)>=1)
nrow(TotQTLs)
```

apaQTL genes:  

```{r}
TotQTLs_gene=TotQTLs %>% group_by(Gene)  %>% summarise(nQTL=n())

summary(TotQTLs_gene$nQTL)

hist(TotQTLs_gene$nQTL)
```


Location distribution for peaks:

```{r}
TotQTLs_loc= TotQTLs %>% group_by(Loc) %>% summarise(nLoc=n()) %>% mutate(PropLoc=nLoc/nrow(TotQTLs))


totQTLloc=ggplot(TotQTLs_loc, aes(x=Loc, y=PropLoc, fill=Loc)) + geom_bar(stat = "Identity") + labs(x="Location of Significant Peak", y="Proportion of QTLs", title="Total QTL peak distribution")+ theme(axis.text.x = element_text(angle = 90, hjust = 1))
```

```{r}
nucRes=read.table("../data/apaQTLPermuted/APApeak_Phenotype_GeneLocAnno.Nuclear_permResBH.txt", stringsAsFactors = F, header = T) %>% separate(pid, into=c("Chr", "Start", "End", "PeakID"), sep=":") %>% separate(PeakID, into=c("Gene", "Loc", "Strand","Peak"), sep="_")
```


Nuclear Apa QTLs   
```{r}
NucQTLs= nucRes %>% filter(-log10(bh)>=1)
nrow(NucQTLs)
```

apaQTL genes:  

```{r}
NucQTLs_gene= NucQTLs %>% group_by(Gene)  %>% summarise(nQTL=n())

summary(NucQTLs_gene$nQTL)

hist(NucQTLs_gene$nQTL)
```

Location distribution for peaks:

```{r}
NucQTLs_loc= NucQTLs %>% group_by(Loc) %>% summarise(nLoc=n()) %>% mutate(PropLoc=nLoc/nrow(NucQTLs))


nucQTLloc=ggplot(NucQTLs_loc, aes(x=Loc, y=PropLoc, fill=Loc)) + geom_bar(stat = "Identity") + labs(x="Location of Significant Peak", y="Proportion of QTLs", title="Nuclear QTL peak distribution")+theme(axis.text.x = element_text(angle = 90, hjust = 1))

```


```{r}
plot_grid(totQTLloc, nucQTLloc)
```

##Distance to PAS  

The distance to PAS is the location of the snp to the end of the peak for the 

Strand in this file is the peak strand (opposite of gene). This means for + strand I want the start of the Peak and for the - strand i will use the end of the peak.
```{r}
TotQTLs_dist=TotQTLs %>% separate(sid, into=c("SnpCHR", "SNPpos"), sep=":") %>% mutate(Dist2PAS=ifelse(Strand=="+", as.integer(SNPpos)-as.integer(Start), as.integer(SNPpos)-as.integer(End)))


summary(abs(TotQTLs_dist$Dist2PAS))
```

Plot:  
```{r}
ggplot(TotQTLs_dist, aes(x=Dist2PAS)) + geom_histogram(bins=100) + theme(axis.text.x = element_text(angle = 90, hjust = 1))
```


```{r}
ggplot(TotQTLs_dist, aes(x=Dist2PAS)) + geom_histogram(bins=100) + theme(axis.text.x = element_text(angle = 90, hjust = 1)) + facet_grid(~Loc)
```
```