Software requirements
Libraries
Quality check
Trim adapters and filter bases based on quality
Discard reads containing protecting group sequence
Alignment
Spike-ins read count analysis
Filtering genomic libraries
Create tdf files
Insert size plots
Calling AP sites
Comparison of SMUG1-snAP-seq sites with peaks obtained in Kawasaki et al., Genome biology, 2017
Merge technical replicates and calculate coverage
Base composition profiles
Coverage profiles centered around the sites
Sequence logos
Exploring TG enrichment using frequencies of dinucleotides counts, motifs and tests

Software requirements

FastQC v0.11.3
cutadapt v1.12
bwa v0.7.15-r1140
samtools v1.3.1
sambamba v0.6.5
Standard Unix tools: awk, sort, uniq
igvtools v2.3.91
deeptools v2.4.2-5-f439d22
bedtools v2.27.0
tableCat.py
EMBOSS v6.6.0.0
fastaRegexFinder.py v0.1.1
meme v4.11.2
macs2 v2.1.1.20160309
python v2.7.12. Libraries:
- os
- collections
- random
- scipy
- numpy
- gzip
- re
- pandas
R v3.3.2. Libraries:

Libraries

ArrayExpress E-MTAB-7152 naming:

Library	Biological replicate	Sequencing type	script id
Lmajor_SMUG1_snAP1	rep1	paired-end	Lib96
Lmajor_SMUG1_snAP2	rep2	paired-end	Lib100
Lmajor_snAP1	rep1	paired-end	Lib97
Lmajor_snAP2	rep2	paired-end	Lib101
Lmajor_input1	rep1	paired-end	Lib99b
Lmajor_input2	rep2	paired-end	Lib103
Lmajor_UNG_snAP1	rep1	paired-end	Lib130
Lmajor_UNG_snAP2	rep2	paired-end	Lib131
Lmajor_UNG_input1	rep1	paired-end	Lib132
Lmajor_UNG_input2	rep2	paired-end	Lib133

Renaming

*_1.fastq.gz -> *_R1.fastq.gz and *_2.fastq.gz -> *_R2.fastq.gz

Quality check

cd ~/fastq # directory containing the fastq sequencing files
mkdir ../fastqc
for fq in *.fastq.gz
do
  fastqc --noextract -q -o ../fastqc $fq
done

Trim adapters and filter bases based on quality

cd ~/fastq
mkdir ../fastq_trimmed

for fq1 in *R1*.fastq.gz
do
  fq2=${fq1/_R1/_R2}
  bname=${fq1%.fastq.gz}
  cutadapt -a AGATCGGAAGAGC -A AGATCGGAAGAGC -m 15 -q 20 -o ../fastq_trimmed/$fq1 -p ../fastq_trimmed/$fq2 $fq1 $fq2 > ../fastq_trimmed/${bname}.txt
done

Discard reads containing protecting group sequence

cd ~/fastq_trimmed
mkdir ../fastq_trimmed_protecting_sequence_optimal

for fq1 in *R1*.fastq.gz
do
  fq2=${fq1/_R1/_R2}
  bname=${fq1%_R1_001.fastq.gz}
  cutadapt -g GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT -G GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCT -e 0.1 -O 20 --discard-trimmed --pair-filter=any -o ../fastq_trimmed_protecting_sequence_optimal/$fq1 -p ../fastq_trimmed_protecting_sequence_optimal/$fq2 $fq1 $fq2 > ../fastq_trimmed_protecting_sequence_optimal/$bname.txt
done

Alignment

Prepare and index references

cd ~/reference/

wget ftp://ftp.sanger.ac.uk/pub/project/pathogens/Leishmania/major/current_gff3/Lmajor.genome.fasta.gz
wget ftp://ftp.sanger.ac.uk/pub/project/pathogens/Leishmania/major/current_gff3/Lmajor.gff3.gz
gzip -d Lmajor.genome.fasta.gz

cat Lmajor.genome.fasta spikeins.fa > Lmajor.genome_spikeins.fasta # see oligo.md for the contents of spikeins.fa

bwa index Lmajor.genome_spikeins.fasta
samtools faidx Lmajor.genome_spikeins.fasta

Align and sort

cd ~/fastq_trimmed_protecting_sequence_optimal

mkdir ../bam

ref=../reference/Lmajor.genome_spikeins.fasta

for fq1 in *R1*.fastq.gz
do
  bname=${fq1%_R1_001.fastq.gz}
  fq2=${fq1/_R1/_R2}
  bwa mem -t 20 -M $ref $fq1 $fq2 | \
  samtools view -@ 20 -b - | \
  samtools sort -@ 20 -T ~/tmp/$bname -o ../bam/$bname.tmp.bam -
done

Merge lanes, mark duplicates, index and flagstat

cd ~/bam

mkdir ../flagstat

listOfIds="..." # add here the list of ids for the libraries you want to merge, if lanes are already merged then skip the samtools merge step below

for id in listOfIds
do
  bams=`echo ${id}_L00[1-4].tmp.bam`
  samtools merge -@ 20 -f ${id}.tmp.bam $bams && \
  sambamba markdup -t 20 ${id}.tmp.bam ${id}.bam 2> ${id}.markdup.txt && \
  samtools flagstat ${id}.bam > ../flagstat/${id}.txt
done

Spike-ins read count analysis

Overall

cd ~/bam

for bam in `ls *.bam | grep -v "tmp"`
do
bname=${bam%.bam}
AP1_fwd=`samtools view -c -F 16 $bam AP1`
AP1_rev=`samtools view -c -f 16 $bam AP1`
GCAT1_fwd=`samtools view -c -F 16 $bam GCAT1`
GCAT1_rev=`samtools view -c -f 16 $bam GCAT1`
fU1_fwd=`samtools view -c -F 16 $bam fU1`
fU1_rev=`samtools view -c -f 16 $bam fU1`
fC1_fwd=`samtools view -c -F 16 $bam fC1`
fC1_rev=`samtools view -c -f 16 $bam fC1`
AP2_fwd=`samtools view -c -F 16 $bam AP2`
AP2_rev=`samtools view -c -f 16 $bam AP2`
GCAT2_fwd=`samtools view -c -F 16 $bam GCAT2`
GCAT2_rev=`samtools view -c -f 16 $bam GCAT2`
fU2_fwd=`samtools view -c -F 16 $bam fU2`
fU2_rev=`samtools view -c -f 16 $bam fU2`
fC2_fwd=`samtools view -c -F 16 $bam fC2`
fC2_rev=`samtools view -c -f 16 $bam fC2`
echo -e "${bname}\t${AP1_fwd}\t${AP1_rev}\t${GCAT1_fwd}\t${GCAT1_rev}\t${fU1_fwd}\t${fU1_rev}\t${fC1_fwd}\t${fC1_rev}\t${AP2_fwd}\t${AP2_rev}\t${GCAT2_fwd}\t${GCAT2_rev}\t${fU2_fwd}\t${fU2_rev}\t${fC2_fwd}\t${fC2_rev}"
done | column -t

All libraries

cd ~/bam

mkdir ../reference_distribution

for bam in `ls *.bam | grep -v "tmp"`
do
  samtools view $bam -F260 -f64 -f32 | cut -f3 | sort | uniq -c | grep -E 'AP1|AP2|fC1|fC2|fU1|fU2|GCAT1|GCAT2' > ../reference_distribution/${bam%.bam}_R1_fwd.txt && \
  samtools view $bam -F260 -f64 -f16 | cut -f3 | sort | uniq -c | grep -E 'AP1|AP2|fC1|fC2|fU1|fU2|GCAT1|GCAT2' > ../reference_distribution/${bam%.bam}_R1_rev.txt && \
  samtools view $bam -F260 -f128 -f32 | cut -f3 | sort | uniq -c | grep -E 'AP1|AP2|fC1|fC2|fU1|fU2|GCAT1|GCAT2' > ../reference_distribution/${bam%.bam}_R2_fwd.txt && \
  samtools view $bam -F260 -f128 -f16 | cut -f3 | sort | uniq -c | grep -E 'AP1|AP2|fC1|fC2|fU1|fU2|GCAT1|GCAT2' > ../reference_distribution/${bam%.bam}_R2_rev.txt
done

Generating a fragmentation pattern file for each library, R1/R2 and fwd/rev.

cd ~/bam

mkdir ../fragmentation_analysis

for bam in `ls *.bam | grep -v "tmp"`
do
  for ref in AP1 AP2 fC1 fC2 fU1 fU2 GCAT1 GCAT2
  do
      nohup samtools view $bam -F260 -f64 -f32 | awk -v ref=$ref '$3==ref {print $4}' | sort -k1,1n | uniq -c | awk '{t = $1; $1 = $2; $2 = t; print; }' > ../fragmentation_analysis/${bam%.bam}_R1_fwd_${ref}.txt &
      nohup samtools view $bam -F260 -f64 -f16 | awk -v ref=$ref '$3==ref {print $4}' | sort -k1,1n | uniq -c | awk '{ t = $1; $1 = $2; $2 = t; print; }' > ../fragmentation_analysis/${bam%.bam}_R1_rev_${ref}.txt &
      nohup samtools view $bam -F260 -f128 -f32 | awk -v ref=$ref '$3==ref {print $4}' | sort -k1,1n | uniq -c | awk '{ t = $1; $1 = $2; $2 = t; print; }' > ../fragmentation_analysis/${bam%.bam}_R2_fwd_${ref}.txt &
      nohup samtools view $bam -F260 -f128 -f16 | awk -v ref=$ref '$3==ref {print $4}' | sort -k1,1n | uniq -c | awk '{ t = $1; $1 = $2; $2 = t; print; }' > ../fragmentation_analysis/${bam%.bam}_R2_rev_${ref}.txt &
    done
done

Filtering genomic libraries

cd ~/bam

for bam in `*.bam | grep -v "tmp"`
do
  bname=${bam%.bam}
  samtools idxstats $bam | cut -f1 | grep -v -E 'AP1|AP2|fC1|fC2|fU1|fU2|GCAT1|GCAT2' | xargs samtools view -@ 20 -F 3844 -q 10 -b $bam > $bname.clean.bam && samtools index $bname.clean.bam && \
  samtools view -@ 20 -f 64 -b $bname.clean.bam > $bname.clean.R1.bam && samtools index $bname.clean.R1.bam && \
  samtools view -@ 20 -f 64 -f 32 -b $bname.clean.bam > $bname.clean.R1.fwd.bam && samtools index $bname.clean.R1.fwd.bam && \
  samtools view -@ 20 -f 64 -f 16 -b $bname.clean.bam > $bname.clean.R1.rev.bam && samtools index $bname.clean.R1.rev.bam
done

Filter R1.fwd and R1.rev by CIGAR

cd ~/bam

#R1.fwd
for bam in *.clean.R1.fwd.bam
do
  bname=${bam%.bam}
  samtools view -@ 20 -h $bam | awk '$6 !~ /^[0-9]+S/' | samtools view -@ 20 -b - > ${bname}.cigar.bam && samtools index ${bname}.cigar.bam
done

#R1.rev
for bam in *.clean.R1.rev.bam
do
  bname=${bam%.bam}
  nohup samtools view -@ 20 -h $bam | awk '$6 !~ /S$/' | samtools view -@ 20 -b - > ${bname}.cigar.bam && samtools index ${bname}.cigar.bam &
done

Create tdf files

cd ~/bam

mkdir ../tdf

#paired-end
for bam in *.clean.bam
do
  bname=${bam%.bam}
  igvtools count $bam ../tdf/$bname.tdf ../reference/Lmajor.genome_spikeins.fasta.fai --includeDuplicates --pairs -w 1 -e 0 --preExtFactor 0 --postExtFactor 0
done

#single-end
for bam in *.clean.R*.bam
do
  bname=${bam%.bam}
  igvtools count $bam ../tdf/$bname.tdf ../reference/Lmajor.genome_spikeins.fasta.fai --includeDuplicates -w 1 -e 0 --preExtFactor 0 --postExtFactor 0
done

Insert size plots

cd ~/bam

mkdir ../stats

for bam in *.clean.bam
do
  bname=${bam%.bam}
  mkdir ../stats/$bname
  samtools stats $bam > ../stats/$bname/$bname.txt && \
  plot-bamstats -p ../stats/$bname/ ../stats/$bname/$bname.txt
done

Calling AP sites

R1.fwd

Coverage of alignment start sites

cd ~/bam

mkdir ../ap

g=../reference/Lmajor.genome_spikeins.fasta.fai

for bam in *.clean.R1.fwd.cigar.bam
do
  bname=`basename ${bam%.bam}`
  bedtools genomecov -d -5 -ibam $bam -g $g | grep 'LmjF*' > ../ap/$bname.cov
done

Obtain sites

library(data.table)
library(edgeR)
library(ggplot2)

# Enlarge the view width when printing tables
options(width = 250)


##########################################
# Lib96 and Lib100 vs. Lib99b and Lib103 #
##########################################

# Load data
data <- fread("tableCat.py -i ~/ap/*Lib{96,100,99b,103}*fwd*.cov -r .clean.R1.fwd.cigar.cov")
setnames(data, c("chr", "pos", "count", "lib"))
table(data$lib)

# Cast
data_cast <- dcast.data.table(data = data, chr + pos ~ lib, value.var = "count")

# Define group
group <- factor(c('ap', 'input', 'ap', 'input'))

# Define DGEList object
y <- DGEList(counts = data_cast[,-c(1,2)], group = group, genes = data_cast[,c(1,2)])
y$samples

# Filter and get the top 100k sites according to cpm
y_f <- y[names(sort(rowMeans(cpm(y)), decreasing=T)[1:1e5]),]

# Define design matrix
des <- model.matrix(~ 0 + group, data = y_f$samples)
colnames(des) <- levels(factor(y_f$samples$group))

# Calculate normalization factors
y_f <- calcNormFactors(y_f, method = "TMM")

# Estimate dispersion
y_f <- estimateDisp(y_f, des)

# Fit linear model
fit <- glmFit(y_f, des, prior.count=1)

# Define contrasts
my.contrasts <- makeContrasts(apVSinput = ap - input, levels = des)

# Obtain likelihoods
lrt <- glmLRT(fit, contrast=my.contrasts[,"apVSinput"])

# Tables, bed files and volcano plots
## tables
detable <- topTags(lrt, n = Inf)$table
detable_e <- data.table(cbind(detable, data.frame(data_cast[,-c(1,2)][as.numeric(row.names(detable)),])))
detable_e[, start := as.integer(pos - 2)]
detable_e[, end := as.integer(start + 1)]
detable_e <- detable_e[,.(chr, start, end, Lib96, Lib100, Lib99b, Lib103, logFC, logCPM, LR, PValue, FDR)]

nrow(detable_e[FDR < 1e-15 & logFC > 0]) # 492 - very high confidence sites
nrow(detable_e[FDR < 1e-15 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-10 & logFC > 0]) # 1560 - high confidence sites
nrow(detable_e[FDR < 1e-10 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-5 & logFC > 0]) # 5635 - sites of confidence
nrow(detable_e[FDR < 1e-5 & logFC < 0]) # 6
nrow(detable_e[FDR < 1e-1 & logFC > 0]) # 38575 - more sites
nrow(detable_e[FDR < 1e-1 & logFC < 0]) # 3354
detable_e[logFC < 0][1,]$FDR # 2.086796e-09 - the best FDR in the opposite direction logFC<0 could be a good threshold for our logFC>0

write.table(detable_e[FDR < 1e-10 & logFC > 0], "~/ap/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.txt", row.names = FALSE, col.names = TRUE, sep = '\t', quote = FALSE)

## bed
write.table(detable_e[FDR < 1e-10 & logFC > 0][, c("chr", "start", "end", "FDR")][start>-1], "~/ap/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.bed", row.names = FALSE, col.names = FALSE, sep = '\t', quote = FALSE)

## volcano
df <- data.table(x = detable_e$logFC, y = -log10(detable_e$FDR), d = densCols(detable_e$logFC, -log10(detable_e$FDR), colramp = colorRampPalette(rev(rainbow(10, end = 4/6)))))

gg <- ggplot(data = df, aes(x, y, col = d)) +
geom_point(size = 0.1) +
scale_color_identity() +
ylab(expression("-log"[10]*"FDR")) +
xlab(expression("log"[2]*"FC")) +
geom_vline(xintercept = 0, linetype = 'dotted') +
theme_bw() +
coord_cartesian(xlim = c(-10, 10))

ggsave('~/figures/Lib96_Leish_SMUG1-AP.Lib100_Leish_SMUG1-AP_2.clean.R1.fwd.cigar.png', width = 12, height = 12, units = 'cm')


##########################################
# Lib97 and Lib101 vs. Lib99b and Lib103 #
##########################################

# Load data
data <- fread("tableCat.py -i ~/ap/*Lib{97,101,99b,103}*fwd*.cov -r .clean.R1.fwd.cigar.cov")
setnames(data, c("chr", "pos", "count", "lib"))
table(data$lib)

# Cast
data_cast <- dcast.data.table(data = data, chr + pos ~ lib, value.var = "count")

# Define group
group <- factor(c('ap', 'input', 'ap', 'input'))

# Define DGEList object
y <- DGEList(counts = data_cast[,-c(1,2)], group = group, genes = data_cast[,c(1,2)])
y$samples

# Filter and get the top 100k sites according to cpm
y_f <- y[names(sort(rowMeans(cpm(y)), decreasing=T)[1:1e5]),]

# Define design matrix
des <- model.matrix(~ 0 + group, data = y_f$samples)
colnames(des) <- levels(factor(y_f$samples$group))

# Calculate normalization factors
y_f <- calcNormFactors(y_f, method = "TMM")

# Estimate dispersion
y_f <- estimateDisp(y_f, des)

# Fit linear model
fit <- glmFit(y_f, des, prior.count=1)

# Define contrasts
my.contrasts <- makeContrasts(apVSinput = ap - input, levels = des)

# Obtain likelihoods
lrt <- glmLRT(fit, contrast=my.contrasts[,"apVSinput"])

# Tables, bed files and volcano plots
## tables
detable <- topTags(lrt, n = Inf)$table
detable_e <- data.table(cbind(detable, data.frame(data_cast[,-c(1,2)][as.numeric(row.names(detable)),])))
detable_e[, start := as.integer(pos - 2)]
detable_e[, end := as.integer(start + 1)]
detable_e <- detable_e[,.(chr, start, end, Lib97, Lib101, Lib99b, Lib103, logFC, logCPM, LR, PValue, FDR)]

nrow(detable_e[FDR < 1e-15 & logFC > 0]) # 0 - very high confidence sites
nrow(detable_e[FDR < 1e-15 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-10 & logFC > 0]) # 0 - high confidence sites
nrow(detable_e[FDR < 1e-10 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-5 & logFC > 0]) # 0 - sites of confidence
nrow(detable_e[FDR < 1e-5 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-1 & logFC > 0]) # 1 - more sites
nrow(detable_e[FDR < 1e-1 & logFC < 0]) # 3
detable_e[logFC < 0][1,]$FDR # 0.01005186

## volcano
df <- data.table(x = detable_e$logFC, y = -log10(detable_e$FDR), d = densCols(detable_e$logFC, -log10(detable_e$FDR), colramp = colorRampPalette(rev(rainbow(10, end = 4/6)))))

gg <- ggplot(data = df, aes(x, y, col = d)) +
geom_point(size = 0.1) +
scale_color_identity() +
ylab(expression("-log"[10]*"FDR")) +
xlab(expression("log"[2]*"FC")) +
geom_vline(xintercept = 0, linetype = 'dotted') +
theme_bw() +
coord_cartesian(xlim = c(-10, 10))

ggsave('~/figures/Lib97_Leish_AP.Lib101_Leish_AP_2.clean.R1.fwd.cigar.png', width = 12, height = 12, units = 'cm')


###########################################
# Lib130 and Lib131 vs. Lib132 and Lib133 #
###########################################

# Load data
data <- fread("tableCat.py -i ~/ap/*Lib13{0..3}*fwd*cov -r .clean.R1.fwd.cigar.cov")
setnames(data, c("chr", "pos", "count", "lib"))
table(data$lib)

# Cast
data_cast <- dcast.data.table(data = data, chr + pos ~ lib, value.var = "count")

# Define group
group <- factor(c('ap', 'ap', 'input', 'input'))

# Define DGEList object
y <- DGEList(counts = data_cast[,-c(1,2)], group = group, genes = data_cast[,c(1,2)])
y$samples

# Filter and get the top 100k sites according to cpm
y_f <- y[names(sort(rowMeans(cpm(y)), decreasing=T)[1:1e5]),]

# Define design matrix
des <- model.matrix(~ 0 + group, data = y_f$samples)
colnames(des) <- levels(factor(y_f$samples$group))

# Calculate normalization factors
y_f <- calcNormFactors(y_f, method = "TMM")

# Estimate dispersion
y_f <- estimateDisp(y_f, des)

# Fit linear model
fit <- glmFit(y_f, des, prior.count=1)

# Define contrasts
my.contrasts <- makeContrasts(apVSinput = ap - input, levels = des)

# Obtain likelihoods
lrt <- glmLRT(fit, contrast=my.contrasts[,"apVSinput"])

# Tables, bed files and volcano plots
## tables
detable <- topTags(lrt, n = Inf)$table
detable_e <- data.table(cbind(detable, data.frame(data_cast[,-c(1,2)][as.numeric(row.names(detable)),])))
detable_e[, start := as.integer(pos - 2)]
detable_e[, end := as.integer(start + 1)]
detable_e <- detable_e[,.(chr, start, end, Lib130, Lib131, Lib132, Lib133, logFC, logCPM, LR, PValue, FDR)]

nrow(detable_e[FDR < 1e-15 & logFC > 0]) # 0 - very high confidence sites
nrow(detable_e[FDR < 1e-15 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-10 & logFC > 0]) # 0 - high confidence sites
nrow(detable_e[FDR < 1e-10 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-5 & logFC > 0]) # 0 - sites of confidence
nrow(detable_e[FDR < 1e-5 & logFC < 0]) # 7
nrow(detable_e[FDR < 1e-1 & logFC > 0]) # 21 - more sites
nrow(detable_e[FDR < 1e-1 & logFC < 0]) # 136
detable_e[logFC < 0][1,]$FDR # 2.588202e-09 - the best FDR in the opposite direction logFC<0 could be a good threshold for our logFC>0

## volcano
df <- data.table(x = detable_e$logFC, y = -log10(detable_e$FDR), d = densCols(detable_e$logFC, -log10(detable_e$FDR), colramp = colorRampPalette(rev(rainbow(10, end = 4/6)))))

gg <- ggplot(data = df, aes(x, y, col = d)) +
geom_point(size = 0.1) +
scale_color_identity() +
ylab(expression("-log"[10]*"FDR")) +
xlab(expression("log"[2]*"FC")) +
geom_vline(xintercept = 0, linetype = 'dotted') +
theme_bw() +
coord_cartesian(xlim = c(-10, 10))

ggsave('~/figures/Lib130_Leish_UNG-AP1.Lib131_Leish_UNG-AP2.clean.R1.fwd.cigar.png', width = 12, height = 12, units = 'cm')

R1.rev

Coverage of alignment start sites

cd ~/bam

g=../reference/Lmajor.genome_spikeins.fasta.fai

for bam in *Leish*.clean.R1.rev.cigar.bam
do
  bname=`basename ${bam%.bam}`
  bedtools genomecov -d -5 -ibam $bam -g $g | grep 'LmjF*' > ../ap/$bname.cov
done

Obtain sites

library(data.table)
library(edgeR)
library(ggplot2)

# Enlarge the view width when printing tables
options(width = 250)


##########################################
# Lib96 and Lib100 vs. Lib99b and Lib103 #
##########################################

# Load data
data <- fread("tableCat.py -i ~/ap/*Lib{96,100,99b,103}*rev*.cov -r .clean.R1.rev.cigar.cov")
setnames(data, c("chr", "pos", "count", "lib"))
table(data$lib)

# Cast
data_cast <- dcast.data.table(data = data, chr + pos ~ lib, value.var = "count")

# Define group
group <- factor(c('ap', 'input', 'ap', 'input'))

# Define DGEList object
y <- DGEList(counts = data_cast[,-c(1,2)], group = group, genes = data_cast[,c(1,2)])
y$samples

# Filter and get the top 100k sites according to cpm
y_f <- y[names(sort(rowMeans(cpm(y)), decreasing=T)[1:1e5]),]

# Define design matrix
des <- model.matrix(~ 0 + group, data = y_f$samples)
colnames(des) <- levels(factor(y_f$samples$group))

# Calculate normalization factors
y_f <- calcNormFactors(y_f, method = "TMM")

# Estimate dispersion
y_f <- estimateDisp(y_f, des)

# Fit linear model
fit <- glmFit(y_f, des, prior.count=1)

# Define contrasts
my.contrasts <- makeContrasts(apVSinput = ap - input, levels = des)

# Obtain likelihoods
lrt <- glmLRT(fit, contrast=my.contrasts[,"apVSinput"])

# Tables, bed files and volcano plots
## tables
detable <- topTags(lrt, n = Inf)$table
detable_e <- data.table(cbind(detable, data.frame(data_cast[,-c(1,2)][as.numeric(row.names(detable)),])))
detable_e[, start := as.integer(pos)]
detable_e[, end := as.integer(start + 1)]
detable_e <- detable_e[,.(chr, start, end, Lib96, Lib100, Lib99b, Lib103, logFC, logCPM, LR, PValue, FDR)]

nrow(detable_e[FDR < 1e-15 & logFC > 0]) # 545 - very high confidence sites
nrow(detable_e[FDR < 1e-15 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-10 & logFC > 0]) # 1640 - high confidence sites
nrow(detable_e[FDR < 1e-10 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-5 & logFC > 0]) # 6081 - sites of confidence
nrow(detable_e[FDR < 1e-5 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-1 & logFC > 0]) # 38994 - more sites
nrow(detable_e[FDR < 1e-1 & logFC < 0]) # 3272
detable_e[logFC < 0][1,]$FDR # 1.365118e-05 - the best FDR in the opposite direction logFC<0 could be a good threshold for our logFC>0

write.table(detable_e[FDR < 1e-10 & logFC > 0], "~/ap/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.rev.cigar_FDR1e-10.txt", row.names = FALSE, col.names = TRUE, sep = '\t', quote = FALSE)

## bed
write.table(detable_e[FDR < 1e-10 & logFC > 0][, c("chr", "start", "end", "FDR")][start>-1], "~/ap/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.rev.cigar_FDR1e-10.bed", row.names = FALSE, col.names = FALSE, sep = '\t', quote = FALSE)

## volcano
df <- data.table(x = detable_e$logFC, y = -log10(detable_e$FDR), d = densCols(detable_e$logFC, -log10(detable_e$FDR), colramp = colorRampPalette(rev(rainbow(10, end = 4/6)))))

gg <- ggplot(data = df, aes(x, y, col = d)) +
geom_point(size = 0.1) +
scale_color_identity() +
ylab(expression("-log"[10]*"FDR")) +
xlab(expression("log"[2]*"FC")) +
geom_vline(xintercept = 0, linetype = 'dotted') +
theme_bw() +
coord_cartesian(xlim = c(-10, 10))

ggsave('~/figures/Lib96_Leish_SMUG1-AP.Lib100_Leish_SMUG1-AP_2.clean.R1.rev.cigar.png', width = 12, height = 12, units = 'cm')


##########################################
# Lib97 and Lib101 vs. Lib99b and Lib103 #
##########################################

# Load data
data <- fread("tableCat.py -i ~/ap/*Lib{97,101,99b,103}*rev*.cov -r .clean.R1.rev.cigar.cov")
setnames(data, c("chr", "pos", "count", "lib"))
table(data$lib)

# Cast
data_cast <- dcast.data.table(data = data, chr + pos ~ lib, value.var = "count")

# Define group
group <- factor(c('ap', 'input', 'ap', 'input'))

# Define DGEList object
y <- DGEList(counts = data_cast[,-c(1,2)], group = group, genes = data_cast[,c(1,2)])
y$samples

# Filter and get the top 100k sites according to cpm
y_f <- y[names(sort(rowMeans(cpm(y)), decreasing=T)[1:1e5]),]

# Define design matrix
des <- model.matrix(~ 0 + group, data = y_f$samples)
colnames(des) <- levels(factor(y_f$samples$group))

# Calculate normalization factors
y_f <- calcNormFactors(y_f, method = "TMM")

# Estimate dispersion
y_f <- estimateDisp(y_f, des)

# Fit linear model
fit <- glmFit(y_f, des, prior.count=1)

# Define contrasts
my.contrasts <- makeContrasts(apVSinput = ap - input, levels = des)

# Obtain likelihoods
lrt <- glmLRT(fit, contrast=my.contrasts[,"apVSinput"])

# Tables, bed files and volcano plots
## tables
detable <- topTags(lrt, n = Inf)$table
detable_e <- data.table(cbind(detable, data.frame(data_cast[,-c(1,2)][as.numeric(row.names(detable)),])))
detable_e[, start := as.integer(pos)]
detable_e[, end := as.integer(start + 1)]
detable_e <- detable_e[,.(chr, start, end, Lib97, Lib101, Lib99b, Lib103, logFC, logCPM, LR, PValue, FDR)]

nrow(detable_e[FDR < 1e-15 & logFC > 0]) # 0 - very high confidence sites
nrow(detable_e[FDR < 1e-15 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-10 & logFC > 0]) # 0 - high confidence sites
nrow(detable_e[FDR < 1e-10 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-5 & logFC > 0]) # 0 - sites of confidence
nrow(detable_e[FDR < 1e-5 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-1 & logFC > 0]) # 0 - more sites
nrow(detable_e[FDR < 1e-1 & logFC < 0]) # 0
detable_e[logFC < 0][1,]$FDR # 0.2735397

## volcano
df <- data.table(x = detable_e$logFC, y = -log10(detable_e$FDR), d = densCols(detable_e$logFC, -log10(detable_e$FDR), colramp = colorRampPalette(rev(rainbow(10, end = 4/6)))))

gg <- ggplot(data = df, aes(x, y, col = d)) +
geom_point(size = 0.1) +
scale_color_identity() +
ylab(expression("-log"[10]*"FDR")) +
xlab(expression("log"[2]*"FC")) +
geom_vline(xintercept = 0, linetype = 'dotted') +
theme_bw() +
coord_cartesian(xlim = c(-10, 10))

ggsave('~/figures/Lib97_Leish_AP.Lib101_Leish_AP_2.clean.R1.rev.cigar.png', width = 12, height = 12, units = 'cm')


###########################################
# Lib130 and Lib131 vs. Lib132 and Lib133 #
###########################################

# Load data
data <- fread("tableCat.py -i ~/ap/*Lib13{0..3}*rev*.cov -r .clean.R1.rev.cigar.cov")
setnames(data, c("chr", "pos", "count", "lib"))
table(data$lib)

# Cast
data_cast <- dcast.data.table(data = data, chr + pos ~ lib, value.var = "count")

# Define group
group <- factor(c('ap', 'ap', 'input', 'input'))

# Define DGEList object
y <- DGEList(counts = data_cast[,-c(1,2)], group = group, genes = data_cast[,c(1,2)])
y$samples

# Filter and get the top 100k sites according to cpm
y_f <- y[names(sort(rowMeans(cpm(y)), decreasing=T)[1:1e5]),]

# Define design matrix
des <- model.matrix(~ 0 + group, data = y_f$samples)
colnames(des) <- levels(factor(y_f$samples$group))

# Calculate normalization factors
y_f <- calcNormFactors(y_f, method = "TMM")

# Estimate dispersion
y_f <- estimateDisp(y_f, des)

# Fit linear model
fit <- glmFit(y_f, des, prior.count=1)

# Define contrasts
my.contrasts <- makeContrasts(apVSinput = ap - input, levels = des)

# Obtain likelihoods
lrt <- glmLRT(fit, contrast=my.contrasts[,"apVSinput"])

# Tables, bed files and volcano plots
## tables
detable <- topTags(lrt, n = Inf)$table
detable_e <- data.table(cbind(detable, data.frame(data_cast[,-c(1,2)][as.numeric(row.names(detable)),])))
detable_e[, start := as.integer(pos)]
detable_e[, end := as.integer(start + 1)]
detable_e <- detable_e[,.(chr, start, end, Lib130, Lib131, Lib132, Lib133, logFC, logCPM, LR, PValue, FDR)]

nrow(detable_e[FDR < 1e-15 & logFC > 0]) # 0 - very high confidence sites
nrow(detable_e[FDR < 1e-15 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-10 & logFC > 0]) # 0 - high confidence sites
nrow(detable_e[FDR < 1e-10 & logFC < 0]) # 0
nrow(detable_e[FDR < 1e-5 & logFC > 0]) # 0 - sites of confidence
nrow(detable_e[FDR < 1e-5 & logFC < 0]) # 4
nrow(detable_e[FDR < 1e-1 & logFC > 0]) # 15 - more sites
nrow(detable_e[FDR < 1e-1 & logFC < 0]) # 66
detable_e[logFC < 0][1,]$FDR # 3.998957e-09 - the best FDR in the opposite direction logFC<0 could be a good threshold for our logFC>0

## volcano
df <- data.table(x = detable_e$logFC, y = -log10(detable_e$FDR), d = densCols(detable_e$logFC, -log10(detable_e$FDR), colramp = colorRampPalette(rev(rainbow(10, end = 4/6)))))

gg <- ggplot(data = df, aes(x, y, col = d)) +
geom_point(size = 0.1) +
scale_color_identity() +
ylab(expression("-log"[10]*"FDR")) +
xlab(expression("log"[2]*"FC")) +
geom_vline(xintercept = 0, linetype = 'dotted') +
theme_bw() +
coord_cartesian(xlim = c(-10, 10))

ggsave('~/figures/Lib130_Leish_UNG-AP1.Lib131_Leish_UNG-AP2.clean.R1.rev.cigar.png', width = 12, height = 12, units = 'cm')

Comparison of SMUG1-snAP-seq sites with peaks obtained in Kawasaki et al., Genome biology, 2017

Processed files:

fk_hmU.bed
fk_hmU_baseJ.bed

Obtain chemseq 5hmU peaks in GSE83384:

fk050_Lmaj_chem1 GSM2200481
fk051_Lmaj_chem2 GSM2200482
fk054_Lmaj_chem3 GSM2200484

Create bed files from the _peaks.txt.gz files

cd ~/bed

zcat GSM2200481_fk050_Lmaj_chem1_peaks.txt.gz | grep "^LmjF.*" | cut -f1-3 > GSM2200481_fk050_Lmaj_chem1_peaks.bed
zcat GSM2200482_fk051_Lmaj_chem2_peaks.txt.gz | grep "^LmjF.*" | cut -f1-3 > GSM2200482_fk051_Lmaj_chem2_peaks.bed
zcat GSM2200484_fk054_Lmaj_chem3_peaks.txt.gz | grep "^LmjF.*" | cut -f1-3 > GSM2200484_fk054_Lmaj_chem3_peaks.bed

wc -l *.bed
#206 GSM2200481_fk050_Lmaj_chem1_peaks.bed
#188 GSM2200482_fk051_Lmaj_chem2_peaks.bed
#145 GSM2200484_fk054_Lmaj_chem3_peaks.bed

Intersection:

cd ~/bed

for bed1 in GSM2200481*.bed fk_*.bed
do
  echo `wc -l $bed1`
  echo `wc -l ../ap/*Lib96*R1.fwd*FDR1e-10.bed`
  echo `wc -l ../ap/*Lib96*R1.rev*FDR1e-10.bed`
  bedtools intersect \
  -sorted \
  -a <(bedtools sort -i $bed1) \
  -b <(cat ../ap/*$id*FDR1e-10.bed | bedtools sort -i -) \
  -wa -u | wc -l
  bedtools intersect \
  -sorted \
  -a <(cat ../ap/*$id*FDR1e-10.bed | bedtools sort -i -) \
  -b <(bedtools sort -i $bed1) \
  -wa -u | wc -l
done

Merge technical replicates and calculate coverage

cd ~/bam

mkdir ../bam_merge
mkdir ../bw_merge

#####################
# Lmajor_SMUG1_snAP #
#####################

# fwd
samtools merge -@ 20 ../bam_merge/Leish_SMUG1-snAP-seq.clean.R1.fwd.cigar.bam Leish_SMUG1-AP_S4.clean.R1.fwd.cigar.bam Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar.bam && \
samtools index ../bam_merge/Leish_SMUG1-snAP-seq.clean.R1.fwd.cigar.bam && \
bamCoverage -b ../bam_merge/Leish_SMUG1-snAP-seq.clean.R1.fwd.cigar.bam -o ../bw_merge/Leish_SMUG1-snAP-seq.clean.R1.fwd.cigar.bw -of bigwig --binSize 1 -p 20 --normalizeUsingRPKM

# rev
samtools merge -@ 20 ../bam_merge/Leish_SMUG1-snAP-seq.clean.R1.rev.cigar.bam Leish_SMUG1-AP_S4.clean.R1.rev.cigar.bam Leish_SMUG1-AP_2_S7.clean.R1.rev.cigar.bam && \
samtools index ../bam_merge/Leish_SMUG1-snAP-seq.clean.R1.rev.cigar.bam && \
bamCoverage -b ../bam_merge/Leish_SMUG1-snAP-seq.clean.R1.rev.cigar.bam -o ../bw_merge/Leish_SMUG1-snAP-seq.clean.R1.rev.cigar.bw -of bigwig --binSize 1 -p 20 --normalizeUsingRPKM

Same for snAP-seq and input-Y libraries.

Base composition profiles

R1.fwd

Generate `.bam` and `.fasta` files

cd ~/ap

mkdir ../base_composition_profiles
mkdir ../base_composition_profiles/bam
mkdir ../base_composition_profiles/fasta
mkdir ../base_composition_profiles/tables

g=../reference/Lmajor.genome_spikeins.fasta

bed=Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.bed

bam=${bed%.clean.R1.fwd.cigar_FDR1e-10.bed}.clean.R1.fwd.cigar.bam
bname=${bed%.bed}

cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 10 | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/$bname.fasta && \
echo "1" && \
samtools view ../bam_merge/$bam -L <(cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 10) -b > ../base_composition_profiles/bam/$bname.inpeaks.bam && \
bedtools bamtobed -i ../base_composition_profiles/bam/$bname.inpeaks.bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 11 -r 0 | bedtools slop -i - -g $g.fai -l 0 -r 10 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/$bname.inpeaks.fasta && \
echo "2" && \
samtools view ../bam_merge/Leish_input_Y.clean.R1.fwd.cigar.bam -L <(cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 10) -b > ../base_composition_profiles/bam/$bname.input_Y.bam && \
bedtools bamtobed -i ../base_composition_profiles/bam/$bname.input_Y.bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 11 -r 0 | bedtools slop -i - -g $g.fai -l 0 -r 10 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/$bname.input_Y.fasta && \
echo "3" && \
samtools view ../bam_merge/$bam -L <(cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 10 | bedtools shuffle -seed 123 -noOverlapping -i - -g $g.fai) -b > ../base_composition_profiles/bam/$bname.shuffle.bam && \
bedtools bamtobed -i ../base_composition_profiles/bam/$bname.shuffle.bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 11 -r 0 | bedtools slop -i - -g $g.fai -l 0 -r 10 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/$bname.shuffle.fasta && \
echo "4" && \
bedtools bamtobed -i ../bam_merge/$bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 11 -r 0 | bedtools slop -i - -g $g.fai -l 0 -r 10 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/${bam%.bam}.fasta

# All reads from Leish_input_Y
bedtools bamtobed -i ../bam_merge/Leish_input_Y.clean.R1.fwd.cigar.bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 11 -r 0 | bedtools slop -i - -g $g.fai -l 0 -r 10 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/Leish_input_Y.clean.R1.fwd.cigar.fasta

Convert fasta file into table

Create tables from the .fasta files:

import os

fasta_files = os.listdir("~/base_composition_profiles/fasta")

for f in fasta_files:
  print f
  ifasta = open("~/base_composition_profiles/fasta/%s" % f, "r")
  ilines = ifasta.read()
  ifasta.close()
  gcat_dict = {}
  for entry in ilines.split(">")[1:]:
    seq = list(entry.split()[1].upper())
    for i in range(0, len(seq)):
      base = seq[i]
      i_m = i - 10
      if base in "ACGT":
        if (i_m, base) in gcat_dict:
          gcat_dict[(i_m, base)]+=1
        else:
          gcat_dict[(i_m, base)]=1
  otable = open("~/base_composition_profiles/tables/%s" % f.replace(".fasta", "")+".txt", "w")
  otable.write("pos\tA\tC\tG\tT\n")
  for pos in range(-10, 11):
    otable.write("%s" % str(pos))
    for base in ["A", "C", "G", "T"]:
      if (pos, base) in gcat_dict:
        otable.write("\t%s" % str(gcat_dict[(pos, base)]))
      else:
        otable.write("\t0")
    otable.write("\n")
  otable.close()

Plotting profiles

library(data.table)
library(ggplot2)

setwd("~/base_composition_profiles/tables")

for (f in list.files(".")){
  print(f)
  # Input data
  data <- fread(f)
  # Calculate percentages
  data <- cbind(data[, "pos"], 100*data[, c("A", "C", "G", "T")]/rowSums(data[, c("A", "C", "G", "T")]))
  # Melt data
  data_melt <- melt(data, id.vars = "pos", variable.name = "base", value.name = "pct")
  # Plot
  gg <- ggplot(data = data_melt, aes(x = pos, y = pct, colour = base)) +
  geom_line(size = 1) +
  #stat_smooth(aes(x = pos, y = pct), method = "lm", formula = y ~ poly(x, 10), se = FALSE) +
  #geom_point() +
  #geom_smooth(span = 0.1, se = FALSE) +
  theme_bw() +
  #theme_classic() +
  ylab("Percentage") +
  xlab("") +
  coord_cartesian(ylim = c(0, 100)) +
  scale_x_continuous(breaks = seq(-10, 10), minor_breaks = NULL) +
  scale_y_continuous(breaks = seq(0, 100, 10), minor_breaks = NULL) +
  theme(axis.text=element_text(size=16), axis.title=element_text(size=24), legend.text=element_text(size=20), legend.title=element_blank())
  # Save
  ggsave(gsub(".txt", ".v2.png", f), width = 20, units = "cm")
}

R1.rev

Generate `.bam` and `.fasta` files

cd ~/ap

g=../reference/Lmajor.genome_spikeins.fasta

bed=Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.rev.cigar_FDR1e-10.bed

bam=${bed%.clean.R1.rev.cigar_FDR1e-10.bed}.clean.R1.rev.cigar.bam
bname=${bed%.bed}

cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 10 | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/$bname.fasta && \
echo "1" && \
samtools view ../bam_merge/$bam -L <(cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 10) -b > ../base_composition_profiles/bam/$bname.inpeaks.bam && \
bedtools bamtobed -i ../base_composition_profiles/bam/$bname.inpeaks.bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 0 -r 11 | bedtools slop -i - -g $g.fai -l 10 -r 0 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/$bname.inpeaks.fasta && \
echo "2" && \
samtools view ../bam_merge/Leish_input_Y.clean.R1.rev.cigar.bam -L <(cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 10) -b > ../base_composition_profiles/bam/$bname.input_Y.bam && \
bedtools bamtobed -i ../base_composition_profiles/bam/$bname.input_Y.bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 0 -r 11 | bedtools slop -i - -g $g.fai -l 10 -r 0 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/$bname.input_Y.fasta && \
echo "3" && \
samtools view ../bam_merge/$bam -L <(cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 10 | bedtools shuffle -seed 123 -noOverlapping -i - -g $g.fai) -b > ../base_composition_profiles/bam/$bname.shuffle.bam && \
bedtools bamtobed -i ../base_composition_profiles/bam/$bname.shuffle.bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 0 -r 11 | bedtools slop -i - -g $g.fai -l 10 -r 0 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/$bname.shuffle.fasta && \
echo "4" && \
bedtools bamtobed -i ../bam_merge/$bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 0 -r 11 | bedtools slop -i - -g $g.fai -l 10 -r 0 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/${bam%.bam}.fasta

# All reads from Leish_input_Y
bedtools bamtobed -i ../bam_merge/Leish_input_Y.clean.R1.rev.cigar.bam | cut -f1-3 | bedtools sort -i - | bedtools flank -i - -g $g.fai -l 0 -r 11 | bedtools slop -i - -g $g.fai -l 10 -r 0 | awk '($3-$2) == 21' | bedtools getfasta -fi $g -bed - > ../base_composition_profiles/fasta/Leish_input_Y.clean.R1.rev.cigar.fasta

Convert fasta file into table

Create tables from the .fasta files:

import os

fasta_files = os.listdir("~/base_composition_profiles/fasta")

for f in fasta_files:
  if "R1.rev" in f:
    print f
    ifasta = open("~/base_composition_profiles/fasta/%s" % f, "r")
    ilines = ifasta.read()
    ifasta.close()
    gcat_dict = {}
    for entry in ilines.split(">")[1:]:
      seq = list(entry.split()[1].upper())
      for i in range(0, len(seq)):
        base = seq[i]
        i_m = i - 10
        if base in "ACGT":
          if (i_m, base) in gcat_dict:
            gcat_dict[(i_m, base)]+=1
          else:
            gcat_dict[(i_m, base)]=1
    otable = open("~/base_composition_profiles/tables/%s" % f.replace(".fasta", "")+".txt", "w")
    otable.write("pos\tA\tC\tG\tT\n")
    for pos in range(-10, 11):
      otable.write("%s" % str(pos))
      for base in ["A", "C", "G", "T"]:
        if (pos, base) in gcat_dict:
          otable.write("\t%s" % str(gcat_dict[(pos, base)]))
        else:
          otable.write("\t0")
      otable.write("\n")
    otable.close()

Plotting profiles

library(data.table)
library(ggplot2)

setwd("~/base_composition_profiles/tables")

for (f in list.files(".")){
  if (grepl("R1.rev", f)){
    print(f)
    # Input data
    data <- fread(f)
    # Calculate percentages
    data <- cbind(data[, "pos"], 100*data[, c("A", "C", "G", "T")]/rowSums(data[, c("A", "C", "G", "T")]))
    # Melt data
    data_melt <- melt(data, id.vars = "pos", variable.name = "base", value.name = "pct")
    # Plot
    gg <- ggplot(data = data_melt, aes(x = pos, y = pct, colour = base)) +
    geom_line(size = 1) +
    #stat_smooth(aes(x = pos, y = pct), method = "lm", formula = y ~ poly(x, 10), se = FALSE) +
    #geom_point() +
    #geom_smooth(span = 0.1, se = FALSE) +
    theme_bw() +
    #theme_classic() +
    ylab("Percentage") +
    xlab("") +
    coord_cartesian(ylim = c(0, 100)) +
    scale_x_continuous(breaks = seq(-10, 10), minor_breaks = NULL) +
    scale_y_continuous(breaks = seq(0, 100, 10), minor_breaks = NULL) +
    theme(axis.text=element_text(size=16), axis.title=element_text(size=24), legend.text=element_text(size=20), legend.title=element_blank())
    # Save
    ggsave(gsub(".txt", ".v2.png", f), width = 20, units = "cm")
  }
}

Coverage profiles centered around the sites

computeMatrix and plotProfile:

cd ~/ap

mkdir ../deeptools

bed_SMUG1_snAP_seq_fwd='Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.bed'
bed_SMUG1-snAP-seq_rev='Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.rev.cigar_FDR1e-10.bed'

bw_SMUG1_snAP_seq_fwd="../bw_merge/Leish_SMUG1-snAP-seq.clean.R1.fwd.cigar.bw"
bw_SMUG1_snAP_seq_rev="../bw_merge/Leish_SMUG1-snAP-seq.clean.R1.rev.cigar.bw"
bw_snAP_seq_fwd="../bw_merge/Leish_snAP-seq.clean.R1.fwd.cigar.bw"
bw_snAP_seq_rev="../bw_merge/Leish_snAP-seq.clean.R1.rev.cigar.bw"
bw_input_Y_fwd="../bw_merge/Leish_input_Y.clean.R1.fwd.cigar.bw"
bw_input_Y_rev="../bw_merge/Leish_input_Y.clean.R1.rev.cigar.bw"


#######################
# reference-point fwd # 1000bp
#######################

## computeMatrix
computeMatrix reference-point \
-R $bed_SMUG1_snAP_seq_fwd \
-S $bw_SMUG1_snAP_seq_fwd $bw_snAP_seq_fwd $bw_input_Y_fwd \
-out ../deeptools/SMUG1_snAP_seq_fwd_referencepoint_bs1_rpkm_1000.mat.gz \
--referencePoint "center" \
-b 1000 \
-a 1000 \
-bs 1 \
--sortRegions no \
--skipZeros \
-p "max"

## plotProfile --plotType "se"
plotProfile \
--matrixFile ../deeptools/SMUG1_snAP_seq_fwd_referencepoint_bs1_rpkm_1000.mat.gz \
-out ../deeptools/SMUG1_snAP_seq_fwd_referencepoint_bs1_rpkm_1000_plotprofile_se.png \
--dpi 300 \
--plotHeight 9 \
--plotWidth 10 \
--plotType "se" \
--colors blue red green \
--refPointLabel "sites" \
--regionsLabel "" \
--samplesLabel "SMUG1-snAP" "snAP" "input" \
-y "RPKM" \
--perGroup


#######################
# reference-point rev # 1000bp
#######################

## computeMatrix
computeMatrix reference-point \
-R $bed_SMUG1_snAP_seq_rev \
-S $bw_SMUG1_snAP_seq_rev $bw_snAP_seq_rev $bw_input_Y_rev \
-out ../deeptools/SMUG1_snAP_seq_fwd_referencepoint_bs1_rpkm_1000.mat.gz \
--referencePoint "center" \
-b 1000 \
-a 1000 \
-bs 1 \
--sortRegions no \
--skipZeros \
-p "max"

## plotProfile --plotType "se"
plotProfile \
--matrixFile ../deeptools/SMUG1_snAP_seq_fwd_referencepoint_bs1_rpkm_1000.mat.gz \
-out ../deeptools/SMUG1_snAP_seq_fwd_referencepoint_bs1_rpkm_1000_plotprofile_se.png \
--dpi 300 \
--plotHeight 9 \
--plotWidth 10 \
--plotType "se" \
--colors blue red green \
--refPointLabel "sites" \
--regionsLabel "" \
--samplesLabel "SMUG1-snAP" "snAP" "input" \
-y "RPKM" \
--perGroup

Sequence logos

Add 5bp flanks and create fasta files

cd ~/ap

mkdir -p ../logos/fasta
mkdir -p ../logos/ggseqlogo

g=../reference/Lmajor.genome_spikeins.fasta

for bed in *.bed
do
  bname=${bed%.bed}
  cat $bed | cut -f1-3 | bedtools sort -i - | bedtools slop -i - -g $g.fai -b 5 | bedtools getfasta -fi $g -bed - > ../logos/fasta/$bname.fasta
done

ggseqlogo

# Load the required packages
require(ggplot2)
require(ggseqlogo)
require(Biostrings)

# Set working directory
setwd("~/logos/fasta")


##########
# R1.fwd #
##########

# Create list with all sequences at once
r1.fwd <- list()
for (f in list.files(pattern = "*R1.fwd*")){
  if (grepl("SMUG1-snAP-seq", f) & grepl("FDR1e-1\\.", f)){r1.fwd[[paste("SMUG1-snAP-seq FDR < 0.1 (n = ",length(as.data.frame(readDNAStringSet(f))$x),")", sep="")]] <- as.data.frame(readDNAStringSet(f))$x[which(sapply(as.data.frame(readDNAStringSet(f))$x, function(x) nchar(x)==11))]}
  if (grepl("SMUG1-snAP-seq", f) & grepl("FDR1e-5\\.", f)){r1.fwd[[paste("SMUG1-snAP-seq FDR < 10^(-5) (n = ",length(as.data.frame(readDNAStringSet(f))$x),")", sep="")]] <- as.data.frame(readDNAStringSet(f))$x[which(sapply(as.data.frame(readDNAStringSet(f))$x, function(x) nchar(x)==11))]}
  if (grepl("SMUG1-snAP-seq", f) & grepl("FDR1e-10\\.", f)){r1.fwd[[paste("SMUG1-snAP-seq FDR < 10^(-10) (n = ",length(as.data.frame(readDNAStringSet(f))$x),")", sep="")]] <- as.data.frame(readDNAStringSet(f))$x[which(sapply(as.data.frame(readDNAStringSet(f))$x, function(x) nchar(x)==11))]}
  if (grepl("SMUG1-snAP-seq", f) & grepl("FDR1e-15\\.", f)){r1.fwd[[paste("SMUG1-snAP-seq FDR < 10^(-15) (n = ",length(as.data.frame(readDNAStringSet(f))$x),")", sep="")]] <- as.data.frame(readDNAStringSet(f))$x[which(sapply(as.data.frame(readDNAStringSet(f))$x, function(x) nchar(x)==11))]}
}

# Create custom colour scheme
cs <- make_col_scheme(chars=c('A', 'C', 'G', 'T'), cols=hue_pal()(4))

# Plot sequence logo bits and prob
gg <- ggplot() + geom_logo(r1.fwd[c(1,4,2,3,5,8,6,7)], method = 'bits', col_scheme = cs) + theme_logo() + facet_wrap(~seq_group, ncol = 4, scales='free_x') + theme(axis.line.y = element_line(color="black"))
ggsave('../figures/SMUG1-snAP-seq.R1.fwd_bits.png', width = 36, height = 12, units = 'cm')

gg <- ggplot() + geom_logo(r1.fwd[c(1,4,2,3,5,8,6,7)], method = 'prob', col_scheme = cs) + theme_logo() + facet_wrap(~seq_group, ncol = 4, scales='free_x') + theme(axis.line.y = element_line(color="black"))
ggsave('../figures/SMUG1-snAP-seq.R1.fwd_prob.png', width = 36, height = 12, units = 'cm')

# e.g. just for SMUG1-snAP-seq FDR < 10^(-10) (n = 1560)
ggseqlogo(r1.fwd[2], method = 'bits', col_scheme = cs)
ggplot() + geom_logo(r1.fwd[2], method = 'bits', col_scheme = cs) + theme_logo()
ggplot() + geom_logo(r1.fwd[2], method = 'bits', col_scheme = cs) + theme_logo() + theme(axis.line.y = element_line(color="black"))
ggplot() + geom_logo(r1.fwd[2], method = 'bits', col_scheme = cs) + theme_logo() + theme(axis.line.x = element_line(color="black"), axis.line.y = element_line(color="black"))


##########
# R1.rev #
##########

# Create list with all sequences at once
r1.rev <- list()
for (f in list.files(pattern = "*R1.rev*")){
  if (grepl("Lib96", f) & grepl("FDR1e-1\\.", f)){r1.rev[[paste("Lib96 and Lib100 FDR < 0.1 (n = ",length(as.data.frame(readDNAStringSet(f))$x),")", sep="")]] <- as.data.frame(readDNAStringSet(f))$x[which(sapply(as.data.frame(readDNAStringSet(f))$x, function(x) nchar(x)==11))]}
  if (grepl("Lib96", f) & grepl("FDR1e-5\\.", f)){r1.rev[[paste("Lib96 and Lib100 FDR < 10^(-5) (n = ",length(as.data.frame(readDNAStringSet(f))$x),")", sep="")]] <- as.data.frame(readDNAStringSet(f))$x[which(sapply(as.data.frame(readDNAStringSet(f))$x, function(x) nchar(x)==11))]}
  if (grepl("Lib96", f) & grepl("FDR1e-10\\.", f)){r1.rev[[paste("Lib96 and Lib100 FDR < 10^(-10) (n = ",length(as.data.frame(readDNAStringSet(f))$x),")", sep="")]] <- as.data.frame(readDNAStringSet(f))$x[which(sapply(as.data.frame(readDNAStringSet(f))$x, function(x) nchar(x)==11))]}
  if (grepl("Lib96", f) & grepl("FDR1e-15\\.", f)){r1.rev[[paste("Lib96 and Lib100 FDR < 10^(-15) (n = ",length(as.data.frame(readDNAStringSet(f))$x),")", sep="")]] <- as.data.frame(readDNAStringSet(f))$x[which(sapply(as.data.frame(readDNAStringSet(f))$x, function(x) nchar(x)==11))]}
}

# Create custom colour scheme
cs <- make_col_scheme(chars=c('A', 'C', 'G', 'T'), cols=hue_pal()(4))

# Plot sequence logo bits and prob
gg <- ggplot() + geom_logo(r1.rev[c(1,4,2,3,5,8,6,7)], method = 'bits', col_scheme = cs) + theme_logo() + facet_wrap(~seq_group, ncol = 4, scales='free_x') + theme(axis.line.y = element_line(color="black"))
ggsave('../figures/SMUG1-snAP-seq.R1.rev_bits.png', width = 36, height = 12, units = 'cm')

gg <- ggplot() + geom_logo(r1.rev[c(1,4,2,3,5,8,6,7)], method = 'prob', col_scheme = cs) + theme_logo() + facet_wrap(~seq_group, ncol = 4, scales='free_x') + theme(axis.line.y = element_line(color="black"))
ggsave('../figures/SMUG1-snAP-seq.R1.rev_prob.png', width = 36, height = 12, units = 'cm')

Exploring TG enrichment using frequencies of dinucleotides counts, motifs and tests

frequencies

Calculate dinucleotide frequencies in the entire Leishmania genome:
Calculate dinucleotide frequencies in Fumi's 5hmU peaks using the Leishmania genome as expected frequencies

cd ~/reference/

compseq Lmajor.genome.fasta -word 2 Lmajor.genome.fasta.comp

#
# Output from 'compseq'
#
# The Expected frequencies are calculated on the (false) assumption that every
# word has equal frequency.
#
# The input sequences are:
#	LmjF.01
#	LmjF.02
#	LmjF.03
#	LmjF.04
#	LmjF.05
#	LmjF.06
#	LmjF.07
#	LmjF.08
#	LmjF.09
#	LmjF.10
# ... et al.
#
#Word size	2
#Total count	32855059
#
#
# Word	Obs Count	Obs Frequency	Exp Frequency	Obs/Exp Frequency
#
#AA	1380153		0.0420073	0.0625000	0.6721171
#AC	1990373		0.0605804	0.0625000	0.9692866
#AG	2073185		0.0631009	0.0625000	1.0096150
#AT	1116719		0.0339893	0.0625000	0.5438281
#CA	2387571		0.0726698	0.0625000	1.1627170
#CC	2363431		0.0719351	0.0625000	1.1509611
#CG	2974892		0.0905459	0.0625000	1.4487349
#CT	2126964		0.0647378	0.0625000	1.0358047
#GA	2053568		0.0625039	0.0625000	1.0000618
#GC	3390395		0.1031925	0.0625000	1.6510797
#GG	2318901		0.0705797	0.0625000	1.1292756
#GT	2005230		0.0610326	0.0625000	0.9765218
#TA	739139		0.0224970	0.0625000	0.3599514
#TC	2108650		0.0641804	0.0625000	1.0268860
#TG	2401122		0.0730823	0.0625000	1.1693162
#TT	1424746		0.0433646	0.0625000	0.6938334
#
#Other	20		0.0000006	0.0000000	10000000000.0000000


cd ~/bed
bedtools getfasta -fi ../reference/Lmajor.genome.fasta -bed fk_hmU.bed > fk_hmU.fasta
samtools faidx fk_hmU.fasta
compseq fk_hmU.fasta -word 2 fk_hmU.fasta.comp -in ~/reference/Lmajor.genome.fasta.comp

#
# Output from 'compseq'
#
# The Expected frequencies are taken from the file: ~/reference/Lmajor.genome.fasta.comp
#
# The input sequences are:
#	256134-259249
#	260820-267767
#	268567-268988
#	263559-269641
#	246789-253447
#	258512-259983
#	383220-384273
#	141-415
#	701-1604
#	125200-132538
# ... et al.
#
#
#Word size	2
#Total count	260140
#
#
# Word	Obs Count	Obs Frequency	Exp Frequency	Obs/Exp Frequency
#
#AA	18324		0.0704390	0.0420073	1.6768275
#AC	13471		0.0517837	0.0605804	0.8547922
#AG	13922		0.0535173	0.0631009	0.8481232
#AT	8277		0.0318175	0.0339893	0.9361029
#CA	16822		0.0646652	0.0726698	0.8898494
#CC	24727		0.0950527	0.0719351	1.3213670
#CG	20519		0.0788768	0.0905459	0.8711246
#CT	13651		0.0524756	0.0647378	0.8105866
#GA	12819		0.0492773	0.0625039	0.7883878
#GC	24967		0.0959752	0.1031925	0.9300603
#GG	25643		0.0985738	0.0705797	1.3966317
#GT	13415		0.0515684	0.0610326	0.8449318
#TA	6010		0.0231029	0.0224970	1.0269345
#TC	12567		0.0483086	0.0641804	0.7527002
#TG	16762		0.0644345	0.0730823	0.8816709
#TT	18242		0.0701238	0.0433646	1.6170743
#
#Other	2		0.0000077	0.0000006	12.8136132

dreme

Create fasta file containing +/- 2bp from FDR 10^(-10) R1.fwd SMUG1-snAP-seq sites
Create negative files:

+/- 2bp from all Ts genome-wide
+/- 2bp from all Ts within 5hmU peaks

cd ~
mkdir tg
cd tg

mkdir -p dreme/bed
mkdir -p dreme/fasta
mkdir -p dreme/run

g=~/reference//Lmajor.genome.fasta

cat ../ap/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.bed | \
cut -f1-3 | \
bedtools sort -i - | \
bedtools slop -i - -g $g.fai -b 2 | \
bedtools getfasta -fi $g -bed - > dreme/fasta/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.fasta

fastaRegexFinder.py -f $g -r T --noreverse | cut -f1-3 | bedtools slop -i - -g $g.fai -b 2 | awk -v OFS="\t" '$3 - $2 == 5' | bedtools getfasta -fi $g -bed - > dreme/fasta/Lmajor.genome.T.2bp.fasta
grep ">" dreme/fasta/Lmajor.genome.T.2bp.fasta | wc -l # 6673621

ref=~/bed/fk_hmU.fasta
fastaRegexFinder.py -f $ref -r T --noreverse | cut -f1-3 | bedtools slop -i - -g $ref.fai -b 2 | awk -v OFS="\t" '$3 - $2 == 5' | bedtools getfasta -fi $ref -bed - > dreme/fasta/fk_hmU.T.2bp.fasta
grep ">" dreme/fasta/fk_hmU.T.2bp.fasta | wc -l # 53505

Run dreme with -k 1-5 against negative files from second step

cd ~/tg

p=dreme/fasta/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.fasta
n=dreme/fasta/Lmajor.genome.T.2bp.fasta

for k in {1..5}
do
  dreme -k $k -norc -oc dreme/run/Lmajor.genome.T.2bp_k$k -p $p -n $n -png
done

p=dreme/fasta/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.fasta
n=dreme/fasta/fk_hmU.T.2bp.fasta

for k in {1..5}
do
  dreme -k $k -norc -oc dreme/run/fk_hmU.T.2bp_k$k -p $p -n $n -png
done

Test for TG enrichment in SMUG1-snAP-seq

Following discussions with Jane:

Calculate TG and TX frequencies in Lib96+Lib100 R1.fwd sites at FDR 10^(-10) (1560)
Generate samples (e.g. 10000) of size 1560 from Ts in Fumi's hmU regions and Ts genome-wide and calculate TG and TX frequencies in each where X = A, C, T
Generate samples (e.g. 10000) of size 1560 from synthetic 5-hmU N-oligo libraries and calculate TG and TX frequencies in each

import collections
import random
from scipy import stats
import numpy
import gzip
import os
import re
import pandas as pd


## Calculate TG and TX frequencies in SMUG1-snAP-seq R1.fwd sites at FDR 10^(-10) (1560)
lib96_lib100_r1_fwd_ifile = open("~/tg/dreme/fasta/Lib96_Leish_SMUG1-AP_S4.Lib100_Leish_SMUG1-AP_2_S7.clean.R1.fwd.cigar_FDR1e-10.fasta", "r")
lib96_lib100_r1_fwd_fasta = lib96_lib100_r1_fwd_ifile.read().split(">")[1:]
lib96_lib100_r1_fwd_ifile.close()

len(lib96_lib100_r1_fwd_fasta) # 1560, good

lib96_lib100_r1_fwd_cnt = collections.Counter()

for i in lib96_lib100_r1_fwd_fasta:
  dn = i.split()[1][2:4]
  lib96_lib100_r1_fwd_cnt[dn] += 1

lib96_lib100_r1_fwd_cnt['tg'] # 1044
lib96_lib100_r1_fwd_cnt['ta'] + lib96_lib100_r1_fwd_cnt['tc'] + lib96_lib100_r1_fwd_cnt['tt'] # 483
lib96_lib100_r1_fwd_cnt['ta'] + lib96_lib100_r1_fwd_cnt['tc'] + lib96_lib100_r1_fwd_cnt['tg'] + lib96_lib100_r1_fwd_cnt['tt'] # 1527
100*float(lib96_lib100_r1_fwd_cnt['tg'])/(lib96_lib100_r1_fwd_cnt['ta'] + lib96_lib100_r1_fwd_cnt['tc'] + lib96_lib100_r1_fwd_cnt['tg'] + lib96_lib100_r1_fwd_cnt['tt']) # 68% (68.36935166994107)
100*float(lib96_lib100_r1_fwd_cnt['ta'] + lib96_lib100_r1_fwd_cnt['tc'] + lib96_lib100_r1_fwd_cnt['tt'])/(lib96_lib100_r1_fwd_cnt['ta'] + lib96_lib100_r1_fwd_cnt['tc'] + lib96_lib100_r1_fwd_cnt['tg'] + lib96_lib100_r1_fwd_cnt['tt']) # 32% (31.63064833005894)



## Generate samples of size 1560 from Ts in hmU regions and calculate TG and TX frequencies in each
fk_hmU_ifile = open("~/tg/dreme/fasta/fk_hmU.T.2bp.fasta", "r")
fk_hmU_fasta = fk_hmU_ifile.read().split(">")[1:]
fk_hmU_ifile.close()

len(fk_hmU_fasta) # 53505, good

fk_hmU_cnt = collections.Counter()

for i in fk_hmU_fasta:
  dn = i.split()[1][2:4]
  fk_hmU_cnt[dn] += 1

fk_hmU_cnt['tg'] # 16743
fk_hmU_cnt['ta'] + fk_hmU_cnt['tc'] + fk_hmU_cnt['tt'] # 36761
fk_hmU_cnt['ta'] + fk_hmU_cnt['tc'] + fk_hmU_cnt['tg'] + fk_hmU_cnt['tt'] # 53504
100*float(fk_hmU_cnt['tg'])/(fk_hmU_cnt['ta'] + fk_hmU_cnt['tc'] + fk_hmU_cnt['tg'] + fk_hmU_cnt['tt']) # 31% (31.29298744019139)
100*float(fk_hmU_cnt['ta'] + fk_hmU_cnt['tc'] + fk_hmU_cnt['tt'])/(fk_hmU_cnt['ta'] + fk_hmU_cnt['tc'] + fk_hmU_cnt['tg'] + fk_hmU_cnt['tt']) # 69% (68.70701255980862)

fk_hmU_s = []

for i in range(1, 10001):
  random.seed(i)
  print(i)
  fk_hmU_fasta_s = random.sample(fk_hmU_fasta, 1560)
  fk_hmU_cnt_s = collections.Counter()
  for i in fk_hmU_fasta_s:
    dn = i.split()[1][2:4]
    fk_hmU_cnt_s[dn] += 1
  tg_pct = 100*float(fk_hmU_cnt_s['tg'])/(fk_hmU_cnt_s['ta'] + fk_hmU_cnt_s['tc'] + fk_hmU_cnt_s['tg'] + fk_hmU_cnt_s['tt'])
  tx_pct = 100*float(fk_hmU_cnt_s['ta'] + fk_hmU_cnt_s['tc'] + fk_hmU_cnt_s['tt'])/(fk_hmU_cnt_s['ta'] + fk_hmU_cnt_s['tc'] + fk_hmU_cnt_s['tg'] + fk_hmU_cnt_s['tt'])
  fk_hmU_s.append((tg_pct, tx_pct))


stats.describe([s[0] for s in fk_hmU_s]) # DescribeResult(nobs=10000, minmax=(27.115384615384617, 35.96153846153846), mean=31.283412596009939, variance=1.3495955748538351, skewness=0.04975492476248911, kurtosis=-0.05909367536955523)
stats.describe([s[1] for s in fk_hmU_s]) # DescribeResult(nobs=10000, minmax=(64.038461538461533, 72.884615384615387), mean=68.716587403990061, variance=1.3495955748538355, skewness=-0.04975492476248899, kurtosis=-0.059093675369556564)
stats.ttest_1samp([s[0] for s in fk_hmU_s], 100*float(lib96_lib100_r1_fwd_cnt['tg'])/(lib96_lib100_r1_fwd_cnt['ta'] + lib96_lib100_r1_fwd_cnt['tc'] + lib96_lib100_r1_fwd_cnt['tg'] + lib96_lib100_r1_fwd_cnt['tt'])) # pvalue=0.0



## Generate samples of size 1560 from Ts genome-wide and calculate TG and TX frequencies in each
lmajor_genome_ifile = open("~/tg/dreme/fasta/Lmajor.genome.T.2bp.fasta", "r")
lmajor_genome_fasta = lmajor_genome_ifile.read().split(">")[1:]
lmajor_genome_ifile.close()

len(lmajor_genome_fasta) # 6673621, good

lmajor_genome_cnt = collections.Counter()

for i in lmajor_genome_fasta:
  dn = i.split()[1][2:4]
  lmajor_genome_cnt[dn] += 1

lmajor_genome_cnt['tg'] # 2401121
lmajor_genome_cnt['ta'] + lmajor_genome_cnt['tc'] + lmajor_genome_cnt['tt'] # 4272497
lmajor_genome_cnt['ta'] + lmajor_genome_cnt['tc'] + lmajor_genome_cnt['tg'] + lmajor_genome_cnt['tt'] # 6673618
100*float(lmajor_genome_cnt['tg'])/(lmajor_genome_cnt['ta'] + lmajor_genome_cnt['tc'] + lmajor_genome_cnt['tg'] + lmajor_genome_cnt['tt']) # 36% (35.97929938453175)
100*float(lmajor_genome_cnt['ta'] + lmajor_genome_cnt['tc'] + lmajor_genome_cnt['tt'])/(lmajor_genome_cnt['ta'] + lmajor_genome_cnt['tc'] + lmajor_genome_cnt['tg'] + lmajor_genome_cnt['tt']) # 64% (64.02070061546826)

lmajor_genome_s = []

for i in range(1, 10001):
  random.seed(i)
  print(i)
  lmajor_genome_fasta_s = random.sample(lmajor_genome_fasta, 1560)
  lmajor_genome_cnt_s = collections.Counter()
  for i in lmajor_genome_fasta_s:
    dn = i.split()[1][2:4]
    lmajor_genome_cnt_s[dn] += 1
  tg_pct = 100*float(lmajor_genome_cnt_s['tg'])/(lmajor_genome_cnt_s['ta'] + lmajor_genome_cnt_s['tc'] + lmajor_genome_cnt_s['tg'] + lmajor_genome_cnt_s['tt'])
  tx_pct = 100*float(lmajor_genome_cnt_s['ta'] + lmajor_genome_cnt_s['tc'] + lmajor_genome_cnt_s['tt'])/(lmajor_genome_cnt_s['ta'] + lmajor_genome_cnt_s['tc'] + lmajor_genome_cnt_s['tg'] + lmajor_genome_cnt_s['tt'])
  lmajor_genome_s.append((tg_pct, tx_pct))


stats.describe([s[0] for s in lmajor_genome_s]) # DescribeResult(nobs=10000, minmax=(30.192307692307693, 40.512820512820511), mean=35.98778068206115, variance=1.4912509477666627, skewness=0.018695718913297632, kurtosis=-0.01373407151315753)
stats.describe([s[1] for s in lmajor_genome_s]) # DescribeResult(nobs=10000, minmax=(59.487179487179489, 69.807692307692307), mean=64.012219317938872, variance=1.4912509477666627, skewness=-0.01869571891335036, kurtosis=-0.013734071513155754)
stats.ttest_1samp([s[0] for s in lmajor_genome_s], 100*float(lib96_lib100_r1_fwd_cnt['tg'])/(lib96_lib100_r1_fwd_cnt['ta'] + lib96_lib100_r1_fwd_cnt['tc'] + lib96_lib100_r1_fwd_cnt['tg'] + lib96_lib100_r1_fwd_cnt['tt'])) # pvalue=0.0



## Generate samples of size 1560 from Lib138 and calculate TG and TX frequencies in each

# Define variables
d = "~/fastq_trimmed"
files = os.listdir(d)
lib = "Lib138"

# Load reads into memory
reads = "".join([gzip.open(d + "/" + f, 'rt').read() for f in files if (lib in f) and ("_R1_001.fastq.gz" in f)]).split("@")[1:]
len(reads) # 594549, ok

# Create dictionary rname : right_barcode
fwd_seq = "GTAGTAGTCGACTAG"

rn_barcode = {}

for r in reads:
  fields = r.split('\n')
  rn = fields[0]
  s = fields[1]
  if fwd_seq in s:
    idx = re.search(fwd_seq, s).start()
    barcode = s[idx-10:idx]
    if len(barcode) == 10 and "N" not in barcode:
      rn_barcode[rn] = barcode


len(rn_barcode.values()) # 407781 barcodes before deduplication
len(set(rn_barcode.values())) # 217029 (53%) unique barcodes when deduplicating

# Deduplicate dictionary
barcode_rn = {v: k for k, v in rn_barcode.iteritems()}
rn_barcode_dedup = {v: k for k, v in barcode_rn.iteritems()}

# Counting like above
lib138_cnt = collections.Counter()

for i in rn_barcode_dedup.values():
  n = i[0]
  lib138_cnt[n] += 1

lib138_cnt['G'] # 86907
lib138_cnt['A'] + lib138_cnt['C'] + lib138_cnt['T'] # 130122
lib138_cnt['A'] + lib138_cnt['C'] + lib138_cnt['G'] + lib138_cnt['T'] # 217029
100*float(lib138_cnt['G'])/(lib138_cnt['A'] + lib138_cnt['C'] + lib138_cnt['G'] + lib138_cnt['T']) # 40% (40.04395725916813)
100*float(lib138_cnt['A'] + lib138_cnt['C'] + lib138_cnt['T'])/(lib138_cnt['A'] + lib138_cnt['C'] + lib138_cnt['G'] + lib138_cnt['T']) # 60% (59.95604274083187)

lib138_s = []

for i in range(1, 10001):
  random.seed(i)
  print(i)
  rn_barcode_dedup_s = random.sample(rn_barcode_dedup.values(), 1560)
  lib138_cnt_s = collections.Counter()
  for i in rn_barcode_dedup_s:
    n = i[0]
    lib138_cnt_s[n] += 1
  tg_pct = 100*float(lib138_cnt_s['G'])/(lib138_cnt_s['A'] + lib138_cnt_s['C'] + lib138_cnt_s['G'] + lib138_cnt_s['T'])
  tx_pct = 100*float(lib138_cnt_s['A'] + lib138_cnt_s['C'] + lib138_cnt_s['T'])/(lib138_cnt_s['A'] + lib138_cnt_s['C'] + lib138_cnt_s['G'] + lib138_cnt_s['T'])
  lib138_s.append((tg_pct, tx_pct))


stats.describe([s[0] for s in lib138_s]) # DescribeResult(nobs=10000, minmax=(35.064102564102562, 44.807692307692307), mean=40.057192307692304, variance=1.5296910162745398, skewness=0.00239954716476055, kurtosis=-0.052475425671872244)
stats.describe([s[1] for s in lib138_s]) # DescribeResult(nobs=10000, minmax=(55.192307692307693, 64.935897435897431), mean=59.942807692307703, variance=1.5296910162745398, skewness=-0.002399547164777818, kurtosis=-0.05247542567187313)



## Generate samples of size 1560 from Lib139 and calculate TG and TX frequencies in each

# Define variables
d = "~/fastq_trimmed"
files = os.listdir(d)
lib = "Lib139"

# Load reads into memory
reads = "".join([gzip.open(d + "/" + f, 'rt').read() for f in files if (lib in f) and ("_R2_001.fastq.gz" in f)]).split("@")[1:]
len(reads) # 594549, ok

# Create dictionaries N10 rname : right_barcode and N21 rname : left_barcode + modified_base + right_barcode
fwd_seq = "GTAGTAGTCGACTAG"

rn_barcode_N10 = {}
rn_barcode_N21_1 = {}
rn_barcode_N21_2 = {}

for r in reads:
  fields = r.split('\n')
  rn = fields[0]
  s = fields[1]
  if fwd_seq in s:
    idx = re.search(fwd_seq, s).start()
    barcode_N10 = s[idx-10:idx]
    barcode_N21 = s[idx-21:idx]
    if len(barcode_N10) == 10 and "N" not in barcode_N10:
      rn_barcode_N10[rn] = barcode_N10
      rn_barcode_N21_1[rn] = barcode_N21
    if len(barcode_N21) == 21 and "N" not in barcode_N21:
      rn_barcode_N21_2[rn] = barcode_N21


len(rn_barcode_N10.values()) # 574272 barcodes before deduplication
len(set(rn_barcode_N10.values())) # 281120 (49%) unique barcodes when deduplicating

len(rn_barcode_N21_1.values()) # 574272 barcodes before deduplication
len(set(rn_barcode_N21_1.values())) # 552203 (96%) unique barcodes when deduplicating

len(rn_barcode_N21_2.values()) # 563264 barcodes before deduplication
len(set(rn_barcode_N21_2.values())) # 552169 (98%) unique barcodes when deduplicating

# Deduplicate N10 dictionary
barcode_rn_N10 = {v: k for k, v in rn_barcode_N10.iteritems()}
rn_barcode_N10_dedup = {v: k for k, v in barcode_rn_N10.iteritems()}
len(set(rn_barcode_N10_dedup.values())) # 281120

# Reduce N21 dictionary based on the deduplication of N10 dictionary
rn_barcode_N21_1_dedup = {rn: rn_barcode_N21_1[rn] for rn in rn_barcode_N10_dedup.keys() if len(rn_barcode_N21_1[rn]) == 21 and "N" not in rn_barcode_N21_1[rn]}
len(rn_barcode_N21_1_dedup) # 276402
len(set(rn_barcode_N21_1_dedup.values())) # 276402

# Deduplicate N21 dictionary
barcode_rn_N21_2 = {v: k for k, v in rn_barcode_N21_2.iteritems()}
rn_barcode_N21_2_dedup = {v: k for k, v in barcode_rn_N21_2.iteritems()}
len(set(rn_barcode_N21_2_dedup.values())) # 552169

# Counting like above
lib139_cnt = collections.Counter()

for i in rn_barcode_N10_dedup.values():
  n = i[0]
  lib139_cnt[n] += 1

lib139_cnt['G'] # 101968
lib139_cnt['A'] + lib139_cnt['C'] + lib139_cnt['T'] # 179152
lib139_cnt['A'] + lib139_cnt['C'] + lib139_cnt['G'] + lib139_cnt['T'] # 281120
100*float(lib139_cnt['G'])/(lib139_cnt['A'] + lib139_cnt['C'] + lib139_cnt['G'] + lib139_cnt['T']) # 36% (36.2720546385885)
100*float(lib139_cnt['A'] + lib139_cnt['C'] + lib139_cnt['T'])/(lib139_cnt['A'] + lib139_cnt['C'] + lib139_cnt['G'] + lib139_cnt['T']) # 64% (63.7279453614115)

lib139_s = []

for i in range(1, 10001):
  random.seed(i)
  print(i)
  rn_barcode_N10_dedup_s = random.sample(rn_barcode_N10_dedup.values(), 1560)
  lib139_cnt_s = collections.Counter()
  for i in rn_barcode_N10_dedup_s:
    n = i[0]
    lib139_cnt_s[n] += 1
  tg_pct = 100*float(lib139_cnt_s['G'])/(lib139_cnt_s['A'] + lib139_cnt_s['C'] + lib139_cnt_s['G'] + lib139_cnt_s['T'])
  tx_pct = 100*float(lib139_cnt_s['A'] + lib139_cnt_s['C'] + lib139_cnt_s['T'])/(lib139_cnt_s['A'] + lib139_cnt_s['C'] + lib139_cnt_s['G'] + lib139_cnt_s['T'])
  lib139_s.append((tg_pct, tx_pct))


stats.describe([s[0] for s in lib139_s]) # DescribeResult(nobs=10000, minmax=(31.153846153846153, 41.53846153846154), mean=36.284782051282058, variance=1.4653650820374611, skewness=-0.010417139801069421, kurtosis=-0.0038983912495167417)
stats.describe([s[1] for s in lib139_s]) # DescribeResult(nobs=10000, minmax=(58.46153846153846, 68.84615384615384), mean=63.71521794871795, variance=1.4653650820374606, skewness=0.010417139801051465, kurtosis=-0.0038983912495158535)
stats.ttest_1samp([s[0] for s in lib139_s], stats.describe([s[0] for s in lib138_s]).mean) # pvalue=0.0
stats.ttest_ind([s[0] for s in lib139_s], [s[0] for s in lib138_s]) # pvalue=0.0



## Generate samples of size 1560 from Lib140 and calculate TG and TX frequencies in each

# Define variables
d = "~/fastq_trimmed"
files = os.listdir(d)
lib = "Lib140"

# Load reads into memory
reads = "".join([gzip.open(d + "/" + f, 'rt').read() for f in files if (lib in f) and ("_R2_001.fastq.gz" in f)]).split("@")[1:]
len(reads) # 839914, ok

# Create dictionaries N10 rname : right_barcode and N21 rname : left_barcode + modified_base + right_barcode
fwd_seq = "GTAGTAGTCGACTAG"

rn_barcode_N10 = {}
rn_barcode_N21_1 = {}
rn_barcode_N21_2 = {}

for r in reads:
  fields = r.split('\n')
  rn = fields[0]
  s = fields[1]
  if fwd_seq in s:
    idx = re.search(fwd_seq, s).start()
    barcode_N10 = s[idx-10:idx]
    barcode_N21 = s[idx-21:idx]
    if len(barcode_N10) == 10 and "N" not in barcode_N10:
      rn_barcode_N10[rn] = barcode_N10
      rn_barcode_N21_1[rn] = barcode_N21
    if len(barcode_N21) == 21 and "N" not in barcode_N21:
      rn_barcode_N21_2[rn] = barcode_N21


len(rn_barcode_N10.values()) # 557467 barcodes before deduplication
len(set(rn_barcode_N10.values())) # 276071 (50%) unique barcodes when deduplicating

len(rn_barcode_N21_1.values()) # 557467 barcodes before deduplication
len(set(rn_barcode_N21_1.values())) # 500449 (89%) unique barcodes when deduplicating

len(rn_barcode_N21_2.values()) # 507145 barcodes before deduplication
len(set(rn_barcode_N21_2.values())) # 500414 (99%) unique barcodes when deduplicating

# Deduplicate N10 dictionary
barcode_rn_N10 = {v: k for k, v in rn_barcode_N10.iteritems()}
rn_barcode_N10_dedup = {v: k for k, v in barcode_rn_N10.iteritems()}
len(set(rn_barcode_N10_dedup.values())) # 276071

# Reduce N21 dictionary based on the deduplication of N10 dictionary
rn_barcode_N21_1_dedup = {rn: rn_barcode_N21_1[rn] for rn in rn_barcode_N10_dedup.keys() if len(rn_barcode_N21_1[rn]) == 21 and "N" not in rn_barcode_N21_1[rn]}
len(rn_barcode_N21_1_dedup) # 252457
len(set(rn_barcode_N21_1_dedup.values())) # 252457

# Deduplicate N21 dictionary
barcode_rn_N21_2 = {v: k for k, v in rn_barcode_N21_2.iteritems()}
rn_barcode_N21_2_dedup = {v: k for k, v in barcode_rn_N21_2.iteritems()}
len(set(rn_barcode_N21_2_dedup.values())) # 500414

# Counting like above
lib140_cnt = collections.Counter()

for i in rn_barcode_N10_dedup.values():
  n = i[0]
  lib140_cnt[n] += 1

lib140_cnt['G'] # 96033
lib140_cnt['A'] + lib140_cnt['C'] + lib140_cnt['T'] # 180038
lib140_cnt['A'] + lib140_cnt['C'] + lib140_cnt['G'] + lib140_cnt['T'] # 276071
100*float(lib140_cnt['G'])/(lib140_cnt['A'] + lib140_cnt['C'] + lib140_cnt['G'] + lib140_cnt['T']) # 35% (34.78561674351888)
100*float(lib140_cnt['A'] + lib140_cnt['C'] + lib140_cnt['T'])/(lib140_cnt['A'] + lib140_cnt['C'] + lib140_cnt['G'] + lib140_cnt['T']) # 65% (65.21438325648113)

lib140_s = []

for i in range(1, 10001):
  random.seed(i)
  print(i)
  rn_barcode_N10_dedup_s = random.sample(rn_barcode_N10_dedup.values(), 1560)
  lib140_cnt_s = collections.Counter()
  for i in rn_barcode_N10_dedup_s:
    n = i[0]
    lib140_cnt_s[n] += 1
  tg_pct = 100*float(lib140_cnt_s['G'])/(lib140_cnt_s['A'] + lib140_cnt_s['C'] + lib140_cnt_s['G'] + lib140_cnt_s['T'])
  tx_pct = 100*float(lib140_cnt_s['A'] + lib140_cnt_s['C'] + lib140_cnt_s['T'])/(lib140_cnt_s['A'] + lib140_cnt_s['C'] + lib140_cnt_s['G'] + lib140_cnt_s['T'])
  lib140_s.append((tg_pct, tx_pct))


stats.describe([s[0] for s in lib140_s]) # DescribeResult(nobs=10000, minmax=(30.256410256410255, 40.0), mean=34.770615384615382, variance=1.4294577610292258, skewness=0.055306571905339115, kurtosis=0.02198226879298204)
stats.describe([s[1] for s in lib140_s]) # DescribeResult(nobs=10000, minmax=(60.0, 69.743589743589737), mean=65.229384615384618, variance=1.4294577610292258, skewness=-0.05530657190533867, kurtosis=0.02198226879298293)
stats.ttest_1samp([s[0] for s in lib140_s], stats.describe([s[0] for s in lib138_s]).mean) # pvalue=0.0
stats.ttest_ind([s[0] for s in lib140_s], [s[0] for s in lib138_s]) # pvalue=0.0
stats.ttest_1samp([s[0] for s in lib140_s], stats.describe([s[0] for s in lib139_s]).mean) # pvalue=0.0
stats.ttest_ind([s[0] for s in lib140_s], [s[0] for s in lib139_s]) # pvalue=0.0



# Is the difference of differences significant?
# Lib96+Lib100 (1)   -   Ts genome-wide (10000)
# Lib138 (mean of 10000)   -   Lib140 (10000)
# Lib139 (mean of 10000)   -   Lib140 (10000)

lib96_lib100_r1_fwd_tg = 100*float(lib96_lib100_r1_fwd_cnt['tg'])/(lib96_lib100_r1_fwd_cnt['ta'] + lib96_lib100_r1_fwd_cnt['tc'] + lib96_lib100_r1_fwd_cnt['tg'] + lib96_lib100_r1_fwd_cnt['tt'])
lib138_tg = stats.describe([s[0] for s in lib138_s]).mean
lib139_tg = stats.describe([s[0] for s in lib139_s]).mean
stats.ttest_ind(lib96_lib100_r1_fwd_tg - numpy.array([s[0] for s in lmajor_genome_s]), lib138_tg - numpy.array([s[0] for s in lib140_s])) # Ttest_indResult(statistic=1585.4222082646975, pvalue=0.0)
stats.ttest_ind(lib96_lib100_r1_fwd_tg / numpy.array([s[0] for s in lmajor_genome_s]), lib138_tg / numpy.array([s[0] for s in lib140_s])) # Ttest_indResult(statistic=985.71895480418505, pvalue=0.0)
stats.ttest_ind(lib96_lib100_r1_fwd_tg - numpy.array([s[0] for s in lmajor_genome_s]), lib139_tg - numpy.array([s[0] for s in lib140_s])) # Ttest_indResult(statistic=1806.159034591137, pvalue=0.0)
stats.ttest_ind(lib96_lib100_r1_fwd_tg / numpy.array([s[0] for s in lmajor_genome_s]), lib139_tg / numpy.array([s[0] for s in lib140_s])) # Ttest_indResult(statistic=1157.5360338049106, pvalue=0.0)
stats.ttest_ind(lib138_tg - numpy.array([s[0] for s in lib140_s]), lib139_tg - numpy.array([s[0] for s in lib140_s])) # Ttest_indResult(statistic=223.10959899938607, pvalue=0.0)
stats.ttest_ind(lib138_tg / numpy.array([s[0] for s in lib140_s]), lib139_tg / numpy.array([s[0] for s in lib140_s])) # Ttest_indResult(statistic=202.64554735085812, pvalue=0.0)
stats.ttest_1samp(lib139_tg / numpy.array([s[0] for s in lib140_s]), 1) # Ttest_1sampResult(statistic=124.44687391129081, pvalue=0.0)



# Output files for visualisation
df = pd.DataFrame({"lib96_lib100_diff":lib96_lib100_r1_fwd_tg - numpy.array([s[0] for s in lmajor_genome_s]), "lib96_lib100_fc":lib96_lib100_r1_fwd_tg / numpy.array([s[0] for s in lmajor_genome_s]), "lib138_diff": lib138_tg - numpy.array([s[0] for s in lib140_s]), "lib138_fc": lib138_tg / numpy.array([s[0] for s in lib140_s]), "lib139_diff": lib139_tg - numpy.array([s[0] for s in lib140_s]), "lib139_fc": lib139_tg / numpy.array([s[0] for s in lib140_s])})
df.to_csv("~/tg/tables/tg_diff_fc.txt", sep='\t', index=False)

Visualisation:

library(ggplot2)
library(data.table)

# Load data
data <- fread("~/tg/tables/tg_diff_fc.txt")

# Split data in diff and fc
data_diff <- data[,grepl("diff", colnames(data)), with=FALSE]
data_fc <- data[,grepl("fc", colnames(data)), with=FALSE]

########
# diff #
########

data_diff_melt <- melt(data_diff, variable.name = "lib", value.name = "diff")

gg <- ggplot(data_diff_melt, aes(x = factor(lib, levels=c("lib96_lib100_diff", "lib138_diff", "lib139_diff")), y = diff)) +
geom_jitter(colour = 'gray', alpha = 0.35, width = 0.3, size = 0.2) +
geom_boxplot(outlier.shape=NA, alpha = 0) +
coord_cartesian(ylim = c(0, 40)) +
ylab("%TG difference") +
xlab("") +
theme_classic() +
theme(axis.title.y=element_text(size=20), axis.text.y=element_text(size=16), axis.text.x=element_text(size=16)) +
scale_x_discrete(labels=c("lib96_lib100_diff" = "Lib96+100\nvs\ngenome-wide", "lib138_diff" = "Lib138\nvs\nLib140", "lib139_diff" = "Lib139\nvs\nLib140"))

ggsave('~/figures/tg_diff.png')

######
# fc #
######

data_fc_melt <- melt(data_fc, variable.name = "lib", value.name = "fc")

gg <- ggplot(data_fc_melt, aes(x = factor(lib, levels=c("lib96_lib100_fc", "lib138_fc", "lib139_fc")), y = fc)) +
geom_hline(yintercept = 1, linetype = "dotted") +
geom_jitter(colour = 'gray', alpha = 0.35, width = 0.3, size = 0.2) +
geom_boxplot(outlier.shape=NA, alpha = 0) +
coord_cartesian(ylim = c(0.5, 2.5)) +
ylab("%TG fold-change") +
xlab("") +
theme_classic() +
theme(axis.title.y=element_text(size=20), axis.text.y=element_text(size=16), axis.text.x=element_text(size=16)) +
scale_x_discrete(labels=c("lib96_lib100_fc" = "Lib96+100\nvs\ngenome-wide", "lib138_fc" = "Lib138\nvs\nLib140", "lib139_fc" = "Lib139\nvs\nLib140"))

ggsave('~/figures/tg_fc.png')

Files

leishmania.md

Latest commit

History

leishmania.md

File metadata and controls

Table of contents

Software requirements

Libraries

Renaming

Quality check

Trim adapters and filter bases based on quality

Discard reads containing protecting group sequence

Alignment

Prepare and index references

Align and sort

Merge lanes, mark duplicates, index and flagstat

Spike-ins read count analysis

Overall

All libraries

Filtering genomic libraries

Filter R1.fwd and R1.rev by CIGAR

Create tdf files

Insert size plots

Calling AP sites

R1.fwd

Coverage of alignment start sites

Obtain sites

R1.rev

Coverage of alignment start sites

Obtain sites

Comparison of SMUG1-snAP-seq sites with peaks obtained in Kawasaki et al., Genome biology, 2017

Merge technical replicates and calculate coverage

Base composition profiles

R1.fwd

Generate .bam and .fasta files

Convert fasta file into table

Plotting profiles

R1.rev

Generate .bam and .fasta files

Convert fasta file into table

Plotting profiles

Coverage profiles centered around the sites

Sequence logos

Add 5bp flanks and create fasta files

ggseqlogo

Exploring TG enrichment using frequencies of dinucleotides counts, motifs and tests

frequencies

dreme

Test for TG enrichment in SMUG1-snAP-seq

Generate `.bam` and `.fasta` files

Generate `.bam` and `.fasta` files