a.content.Rmd

---
title: "“A” content in each mapped read"
author: "Briana Mittleman"
date: "6/11/2018"
output: html_document
---

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

The goal of this analysis is to start to understand the sequence composition of the three prime seq reads. This may help me detect misspriming at AAAAA rich regions rather than true site usage. The genomic sequence does not carry the polyadenylation signal, this means reads mapping to a genomic AAAAAA region may be false positives.  Gruber et al. removed reads that consisted of more than 80% AAAA. 


##Percent nucleotide in bins  

One method is to measure the number of AAAAAs in my bins with bedtools nuc. I willl need a fasta file and a bed file with the bin.  

```{r}
library(workflowr)
library(ggplot2)
library(dplyr)
library(cowplot)
library(tidyr)
library(reshape2)

```





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

#SBATCH --job-name=nuc.bin
#SBATCH --time=8:00:00
#SBATCH --output=nuc.bin.out
#SBATCH --error=nuc.bin.err
#SBATCH --partition=broadwl
#SBATCH --mem=20G
#SBATCH --mail-type=END

module load Anaconda3  

source activate three-prime-env


bedtools nuc -s -fi /project2/gilad/briana/genome_anotation_data/genome/Homo_sapiens.GRCh37.75.dna_sm.all.fa -bed /project2/gilad/briana/genome_anotation_data/an.int.genome_200_strandspec.bed > /project2/gilad/briana/threeprimeseq/data/bin200.nuccov.bed  
```

I can now pull this file into R.  
```{r}
bin_nuccov=read.table("../data/bin200.nuccov.bed")
names(bin_nuccov)=c("chr", "start", "end", "bin", "score", "strand", "gene", "pct_at", "pct_gc", "numA", "numC", "numG", "numT", "numN", "numOther", "seqlen")

perc_A_bin=bin_nuccov %>% select("chr", "start", "end","bin", "strand", "gene", "numA") %>% mutate(percA=numA/200)
```



```{r}
ggplot(perc_A_bin, aes(percA)) + geom_histogram(bins=30) + labs(title="Percent of As in the bins")
```


I will apply the same filter as I did in the cov.200bp.wind file. I will keep bins with greater than 0 reads in half of the libraries.  

```{r}
cov_all=read.table("../data/ssFC200.cov.bed", header = T, stringsAsFactors = FALSE)
#remember name switch!
names=c("Geneid","Chr", "Start", "End", "Strand", "Length", "N_18486","T_18486","N_18497","T_18497","N_18500","T_18500","N_18505",'T_18505',"N_18508","T_18508","N_18853","T_18853","N_18870","T_18870","N_19128","T_19128","N_19141","T_19141","N_19193","T_19193","N_19209","T_19209","N_19223","N_19225","T_19225","T_19223","N_19238","T_19238","N_19239","T_19239","N_19257","T_19257")

colnames(cov_all)= names

cov_nums_only=cov_all[,7:38]

keep.exprs=rowSums(cov_nums_only>0) >= 16

cov_all_filt=cov_all[keep.exprs,] 

cov_all_filt_bins= cov_all_filt %>% separate(col=Geneid, into=c("bin","gene"), sep=".E") %>% select(bin)

cov_all_filt_bins$bin=as.integer(cov_all_filt_bins$bin)
```


I will intersect the percA file with the bins in the filtered file.  

```{r}
perc_A_bin_filt= perc_A_bin %>% semi_join(cov_all_filt_bins, by="bin")
```
I can no plot the distribution of percA in this.  

```{r}
ggplot(perc_A_bin_filt, aes(percA)) + geom_histogram(bins = 30) + labs(title="Percent of As in the bins after filtering")
```

I will compare this distribution to those for other nucleotides.  (C)  


```{r}

perc_C_bin=bin_nuccov %>% select("chr", "start", "end","bin", "strand", "gene", "numC") %>% mutate(percC=numC/200)

ggplot(perc_C_bin, aes(percC)) + geom_histogram(bins=30) + labs(title="Percent of Cs in the bins")

perc_C_bin_filt= perc_C_bin %>% semi_join(cov_all_filt_bins, by="bin")
ggplot(perc_C_bin_filt, aes(percC)) + geom_histogram(bins = 30) + labs(title="Percent of Cs in the bins after filtering")
```
'
For G  


```{r}

perc_G_bin=bin_nuccov %>% select("chr", "start", "end","bin", "strand", "gene", "numG") %>% mutate(percG=numG/200)

ggplot(perc_G_bin, aes(percG)) + geom_histogram(bins=30) + labs(title="Percent of Gs in the bins")

perc_G_bin_filt= perc_G_bin %>% semi_join(cov_all_filt_bins, by="bin")
ggplot(perc_G_bin_filt, aes(percG)) + geom_histogram(bins = 30) + labs(title="Percent of Gs in the bins after filtering")
```

for T  


```{r}

perc_T_bin=bin_nuccov %>% select("chr", "start", "end","bin", "strand", "gene", "numT") %>% mutate(percT=numT/200)

ggplot(perc_T_bin, aes(percT)) + geom_histogram(bins=30) + labs(title="Percent of Ts in the bins")

perc_T_bin_filt= perc_T_bin %>% semi_join(cov_all_filt_bins, by="bin")
ggplot(perc_T_bin_filt, aes(percT)) + geom_histogram(bins = 30) + labs(title="Percent of Ts in the bins after filtering")
```

Now I will join all of the percent usage of each nucleotide in the filtered bins so I can plot them on one plot.  

```{r}
percNuc= perc_A_bin_filt %>% left_join(perc_T_bin_filt, by=c("chr", "start", "end", "bin", "strand", "gene")) %>% left_join(perc_G_bin_filt, by=c("chr", "start", "end", "bin", "strand", "gene")) %>% left_join(perc_C_bin_filt, by=c("chr", "start", "end", "bin", "strand", "gene")) %>% select("bin", "percA", "percT", "percG", "percC")

percNuc_melt=melt(percNuc, id.vars = "bin")

ggplot(percNuc_melt, aes(value)) + geom_histogram(bins = 30) + facet_wrap(~variable) + labs(title="Percent each nucleotide in filtered bins")
```


##Strech of Nucleotides   
Next check is if the bins have 5 As in a row. I can do this using bedtools nuc as well.  
###Five A's  

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

#SBATCH --job-name=nucA
#SBATCH --time=8:00:00
#SBATCH --output=nucA.out
#SBATCH --error=nucA.err
#SBATCH --partition=broadwl
#SBATCH --mem=20G
#SBATCH --mail-type=END

module load Anaconda3  

source activate three-prime-env


bedtools nuc -s -fi /project2/gilad/briana/genome_anotation_data/genome/Homo_sapiens.GRCh37.75.dna_sm.all.fa -bed /project2/gilad/briana/genome_anotation_data/an.int.genome_200_strandspec.bed -pattern "AAAAA" > /project2/gilad/briana/threeprimeseq/data/bin200.5.A.nuccov.bed  
```


```{r}
bin_Anuccov=read.table("../data/bin200.Anuccov.bed")
names(bin_Anuccov)=c("chr", "start", "end", "bin", "score", "strand", "gene", "pct_at", "pct_gc", "numA", "numC", "numG", "numT", "numN", "numOther", "seqlen", "fiveA")

hist(bin_Anuccov$fiveA)
```


I will filter this the same way I filtered the other file.  

```{r}
 bin_Anuccov_filt = bin_Anuccov %>% semi_join(cov_all_filt_bins, by="bin") %>% select( bin,  gene,  fiveA)

hist(bin_Anuccov_filt$fiveA)
summary(bin_Anuccov_filt$fiveA)
```
Count the number of bins with each value for number of 5 AAAAA  

```{r}
countA_regions= bin_Anuccov_filt %>% group_by(fiveA) %>% count(fiveA)
```

###Five T's  

I will compare this to regions with 5 T's  
(This isnt the best comparison because the probability of each of these stretches genome wide may be different)  


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

#SBATCH --job-name=nucT
#SBATCH --time=8:00:00
#SBATCH --output=nucT.out
#SBATCH --error=nucT.err
#SBATCH --partition=broadwl
#SBATCH --mem=20G
#SBATCH --mail-type=END

module load Anaconda3  

source activate three-prime-env


bedtools nuc -s -fi /project2/gilad/briana/genome_anotation_data/genome/Homo_sapiens.GRCh37.75.dna_sm.all.fa -bed /project2/gilad/briana/genome_anotation_data/an.int.genome_200_strandspec.bed -pattern "TTTTT" > /project2/gilad/briana/threeprimeseq/data/bin200.5.T.nuccov.bed  
```


```{r}
bin_Tnuccov=read.table("../data/bin200.5.T.nuccov.bed")
names(bin_Tnuccov)=c("chr", "start", "end", "bin", "score", "strand", "gene", "pct_at", "pct_gc", "numA", "numC", "numG", "numT", "numN", "numOther", "seqlen", "fiveT")

summary(bin_Tnuccov$fiveT)

bin_Tnuccov_filt = bin_Tnuccov %>% semi_join(cov_all_filt_bins, by="bin") %>% select( bin,  gene,  fiveT)

hist(bin_Tnuccov_filt$fiveT)
summary(bin_Tnuccov_filt$fiveT)


countT_regions= bin_Tnuccov_filt %>% group_by(fiveT) %>% count(fiveT)

```
The numbers are not super different.  

##Correlation between count and A statistics  
  
####Percent nucleotide   
cov_all_filt

```{r}
#select bin coverage fore 18486
cov_all_filt_18486=cov_all_filt %>%separate(col=Geneid, into=c("bin","gene"), sep=".E") %>% select(bin, T_18486)
cov_all_filt_18486$bin=as.integer(cov_all_filt_18486$bin)
#join with percA
perc_A_bin_filt_cov= perc_A_bin_filt %>% select(bin, percA) %>% right_join(cov_all_filt_18486,by="bin") 
#melt it 
perc_A_bin_filt_cov_melt=melt(perc_A_bin_filt_cov, id.vars="bin") 

#plot it 

perA_18486total=ggplot(perc_A_bin_filt_cov, aes(y=T_18486, x=percA)) + geom_point()+ labs(y="Total Bin count",x="Percent A in bin") + geom_smooth(method="lm", col="red")



#select bin coverage fore 18486
cov_all_filt_18486N=cov_all_filt %>%separate(col=Geneid, into=c("bin","gene"), sep=".E") %>% select(bin, N_18486)
cov_all_filt_18486N$bin=as.integer(cov_all_filt_18486N$bin)
#join with percA
perc_A_bin_filt_covN= perc_A_bin_filt %>% select(bin, percA) %>% right_join(cov_all_filt_18486N,by="bin") 
#melt it 
perc_A_bin_filt_cov_melNt=melt(perc_A_bin_filt_covN, id.vars="bin") 
perA_18486nuc=ggplot(perc_A_bin_filt_covN, aes(y=N_18486, x=percA)) + geom_point() + labs(y="Nuclear Bin count",x="Percent A in bin") + geom_smooth(method="lm", col="red")
title <- ggdraw() + draw_label("No relationship between bin read count and percentage \n of A nucleotides in a bin ", fontface = 'bold')

x=plot_grid(perA_18486total,perA_18486nuc)

grid.plot=plot_grid(title, x,ncol=1, rel_heights=c(0.3, 1))

ggsave( "../output/plots/perc.A.bincount.png", grid.plot, width = 10, height = 7)
```

```{r}

lm(perc_A_bin_filt_covN$percA~perc_A_bin_filt_covN$N_18486 )

lm( perc_A_bin_filt_cov$percA ~ perc_A_bin_filt_cov$T_18486)
```

Both have about a 0 correlation. 


I will remove bins we did not have coverage in.  

```{r}
perc_A_bin_filt_cov_no0= perc_A_bin_filt %>% select(bin, percA) %>% right_join(cov_all_filt_18486,by="bin") %>% filter(T_18486 >0 )

ggplot(perc_A_bin_filt_cov_no0, aes(y=T_18486, x=percA)) + geom_point() + geom_smooth(method="lm", col="red")
```

Check for T  

```{r}


#join with percA
perc_T_bin_filt_cov= perc_T_bin_filt %>% select(bin, percT) %>% right_join(cov_all_filt_18486,by="bin") 
#melt it 
perc_T_bin_filt_cov_melt=melt(perc_T_bin_filt_cov, id.vars="bin") 

#plot it 

ggplot(perc_T_bin_filt_cov, aes(y=T_18486, x=percT)) + geom_point()

```


####Number of times we see 5 of a nucleotide  

```{r}

#join with fiveA
bin_Anuccov_filt_cov= bin_Anuccov_filt %>% select(bin, fiveA) %>% right_join(cov_all_filt_18486,by="bin") 


#plot it 

ggplot(bin_Anuccov_filt_cov, aes(y=T_18486, x=fiveA)) + geom_point()

```

It does not look like these are drivers of the variation we see in counts.  


This analysis has shown me that mispriming is not a global problem in the samples. We do not see a correlaation bertween percent As in a bin and the read count for either the total or the nuclear. I also do not see any outliers bins with high coverage and high A percentage. If this is a problem it is at a specfici locus level. We can assess this in regions we see differences between total and nuclear fractions.