mispriming_approach.Rmd

---
title: "Mispriming Approach"
author: "Briana Mittleman"
date: "1/11/2019"
output: html_document
---

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

In this analysis I am gonig to explore the ways to handle mispriming in the 3' seq data. Some people call this internal priming. This is when the polyDt primer attached to an RNA molecule that has a long stretch of A's rather than to the tail. You can identify when this is happening because polyA tails are not in the genome but mispriming As are. In my data I need to look for Ts upstream of the read. This is because our reads are on the opposite strand.

Sheppard et al. cited 2 other papers, Beaudoing et al 2000 and Tian et al 2005. Thet excluded reads with 6 consequitive upstream As or those with 7 in a 10nt window. They did this at the read level.  


I started thinking about this in https://brimittleman.github.io/threeprimeseq/filter_As.html. I did not have it mapped out correctly because I was looking for A's on one strand and T's on the other.  


I will assess the problem then will create a blacklist to get rid of the reads. I should do this in the snakefile before we create BW for the peak calling. 

I can start by updating 6up_bed.sh. To make a new script that grabs the upstream 10 bases. I will look for7 of 10 T's in this region. I am going to do this in python because it is more straight forward to read then an awk script. I can also wrap it easier this way. I can also account for negative values and values larger than the chromosome this way. 


##Retreive 10 upstream bases for each read

Upstream10Bases.py
```{bash,eval=F}
#python  
def main(Fin, Fout):
  outBed=open(Fout, "w")
  chrom_lengths=open("/project2/gilad/briana/genome_anotation_data/chrom_lengths2.sort.bed","r")
  #make a dictionary with chrom lengths
  length_dic={}
  for i in chrom_lengths:
    chrom, start, end = i.split()
    length_dic[str(chrom)]=int(end)  

#write file 
  for ln in open(Fin):
    chrom, start, end, name, score, strand = ln.split()
    chrom=str(chrom)
    if strand=="+":
      start_new=int(start)-10
      if start_new <= 1:
        start_new = 0 
      end_new= int(start)
      if end_new == 0:
        end_new=1
      outBed.write("%s\t%d\t%d\t%s\t%s\t%s\n"%(chrom, start_new, end_new, name, score, strand))
    if strand == "-":
      start_new=int(end)
      end_new=int(end) + 10
      if end_new >= length_dic[chrom]:
        end_new = length_dic[chrom]
        start_new=end_new-1
      outBed.write("%s\t%d\t%d\t%s\t%s\t%s\n"%(chrom, start_new, end_new, name, score, strand))
  outBed.close()  

if __name__ == "__main__":
    import sys
    inFile = sys.argv[1]
    fileNoPath=inFile.split("/")[-1]
    fileshort=fileNoPath[:-4]
    outFile="/project2/gilad/briana/threeprimeseq/data/bed_10up/" + fileshort + "10up.bed"
    main(inFile, outFile)


```

I can wrap this for all of the files.  
wrap_Upstream10Bases.sh
```{bash,eval=F}
#!/bin/bash

#SBATCH --job-name=w_Upstream10Bases
#SBATCH --account=pi-yangili1
#SBATCH --time=8:00:00
#SBATCH --output=w_Upstream10Bases.out
#SBATCH --error=w_Upstream10Bases.err
#SBATCH --partition=broadwl
#SBATCH --mem=8G
#SBATCH --mail-type=END


module load Anaconda3  

source activate three-prime-env


for i in $(ls /project2/gilad/briana/threeprimeseq/data/bed_sort/*-combined-sort.bed); do
            python  Upstream10Bases.py  $i 
        done

```

I need to sort the files:  


Next step is running the nuc function to get the sequences of the positions I just put in the bed files.  



bedtools nuc  

* -fi (fasta file) /project2/gilad/briana/genome_anotation_data/genome/Homo_sapiens.GRCh37.75.dna_sm.all.fa  

* -bed results from 10up stream  

* -s strand specific  

* -seq print exracted sequence  

* > output  


Nuc10BasesUp.sh  
```{bash,eval=F}
#!/bin/bash

#SBATCH --job-name=Nuc10BasesUp
#SBATCH --account=pi-yangili1
#SBATCH --time=8:00:00
#SBATCH --output=Nuc10BasesUp.out
#SBATCH --error=Nuc10BasesUp.err
#SBATCH --partition=broadwl
#SBATCH --mem=36G
#SBATCH --mail-type=END

for i in $(ls /project2/gilad/briana/threeprimeseq/data/bed_10up/*);do
   describer=$(echo ${i} | sed -e 's/.*YL-SP-//' | sed -e "s/-combined-sort10up.bed$//")
   bedtools nuc -s -seq -fi /project2/gilad/briana/genome_anotation_data/genome/Homo_sapiens.GRCh37.75.dna_sm.all.fa -bed $i > /project2/gilad/briana/threeprimeseq/data/nuc_10up/TenBaseUP.${describer}.txt   
done
```

##Evaluate problem in 1 line  


```{r}
library(data.table)
require(ggseqlogo)
library(workflowr)
library(tidyverse)
```

###Nuclear 
Goals for this section:  


* make a seq logo plot 

I made logo plot in https://brimittleman.github.io/Net-seq/explore_umi_usage.html with ggseq logo. 

	        
	        
```{r}
res_colNames=c("chrom","start", "end", "name", "score", "strand", "pctAT", "pctGC", "A", "C", "G", "T", "N", "Other", "Length", "Seq")
```


```{r}

nuc_18486_N= fread("../data/nuc_10up/TenBaseUP.18486-N.txt", col.names = res_colNames)
```
Extract seq for seq logo plot:  

```{r}
#filter for full 10 bp  - removes 422 reads (too close to ends)
nuc_18486_N=nuc_18486_N %>% filter(Length==10)
seqs_18486N= nuc_18486_N$Seq
```
Scheme for logo plot:  


```{r}
cs1 = make_col_scheme(chars=c('A', 'T', 'C', 'G', 'N'), groups=c('A', 'T', 'C', 'G', 'N'), cols=c('red', 'blue', 'green', 'yellow', 'pink'))
```
Create plot:  

```{r}
ggseqlogo(seqs_18486N, col_scheme=cs1,  method = 'prob')
```

This is not overwhelming:  


* count number of non passing reads with 6 T's in a row

```{r}
SixT="TTTTTT"
nuc_18486_N_6Ts=nuc_18486_N %>% filter(grepl(SixT, Seq)) 

perc_Bad6T= nrow(nuc_18486_N_6Ts)/nrow(nuc_18486_N)
perc_Bad6T
```


* count number of non passing with 7 of 10 T's


```{r}
nuc_18486_N_70perc= nuc_18486_N%>% mutate(percT=T/Length) %>% filter(percT>=.7) 

perc_Bad70= nrow(nuc_18486_N_70perc)/nrow(nuc_18486_N)

perc_Bad70
```



* count number of total non passing reads  

For this I need to use an or statement.  

```{r}
nuc_18486_N_bad=  nuc_18486_N%>% mutate(percT=T/Length) %>% filter(percT>=.7 |grepl(SixT, Seq) )

perc_Bad=nrow(nuc_18486_N_bad)/nrow(nuc_18486_N)

perc_Bad
```
This shows us that 46% of reads pass these filters.  

###Total  

```{r}
nuc_18486_T= fread("../data/nuc_10up/TenBaseUP.18486-T.txt", col.names = res_colNames)
``` 

* Seqlogo plot  

Filter less than 10 base pair in length for seqlogo  

```{r}
nuc_18486_T=nuc_18486_T %>% filter(Length==10)
seqs_18486T= nuc_18486_T$Seq
```


Create plot:  

```{r}
ggseqlogo(seqs_18486T, col_scheme=cs1,  method = 'prob')
```



* count number of non passing reads with 6 T's in a row

```{r}

nuc_18486_T_6Ts=nuc_18486_T %>% filter(grepl(SixT, Seq)) 

perc_Bad6T_tot= nrow(nuc_18486_T_6Ts)/nrow(nuc_18486_T)
perc_Bad6T_tot
```


* count number of non passing with 7 of 10 T's


```{r}
nuc_18486_T_70perc= nuc_18486_T%>% mutate(percT=T/Length) %>% filter(percT>=.7) 

perc_Bad70_tot= nrow(nuc_18486_T_70perc)/nrow(nuc_18486_T)

perc_Bad70_tot
```



* count number of total non passing reads  

For this I need to use an or statement.  

```{r}
nuc_18486_T_bad=  nuc_18486_T%>% mutate(percT=T/Length) %>% filter(percT>=.7 |grepl(SixT, Seq) )

perc_Bad_tot=nrow(nuc_18486_T_bad)/nrow(nuc_18486_T)

perc_Bad_tot
```
This shows us that 25% of reads pass these filters   



##For all
I may have to use python when i look at all beacuse this is not fast.

##Extra not using

sort_10upbedFile.sh 
```{bash,eval=F}
#!/bin/bash

#SBATCH --job-name=sort_10upbedFile
#SBATCH --account=pi-yangili1
#SBATCH --time=8:00:00
#SBATCH --output=sort_10upbedFile.out
#SBATCH --error=sort_10upbedFile.err
#SBATCH --partition=broadwl
#SBATCH --mem=8G
#SBATCH --mail-type=END

for i in $( ls /project2/gilad/briana/threeprimeseq/data/bed_10up/*);do
  describer=$(echo ${i} | sed -e 's/.*YL-SP-//' | sed -e "s/-combined-sort10up.bed$//")
  sort -k 1,1 -k2,2n $i > /project2/gilad/briana/threeprimeseq/data/bed_10up_sort/YL-SP-${describer}-combined-sort10up.sort.bed
  done
```