Recreating (and maybe improving on) some of the figures generated with plot-bamstats application in R.
First lets load knitr, tidyverse, reshape2 and gridExtra packages.
library(knitr)
library(tidyverse)
## ── Attaching packages ──────────────────────────────────────────────────────────────────────── tidyverse 1.2.1 ──
## ✔ ggplot2 3.0.0 ✔ purrr 0.2.5
## ✔ tibble 1.4.2 ✔ dplyr 0.7.6
## ✔ tidyr 0.8.1 ✔ stringr 1.3.1
## ✔ readr 1.1.1 ✔ forcats 0.3.0
## ── Conflicts ─────────────────────────────────────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag() masks stats::lag()
library(reshape2)
##
## Attaching package: 'reshape2'
## The following object is masked from 'package:tidyr':
##
## smiths
library(gridExtra)
##
## Attaching package: 'gridExtra'
## The following object is masked from 'package:dplyr':
##
## combine
This document assumes you have the file ‘bwa.samtools.stats’ in your current working directory, lets test to make sure it is.
getwd()
## [1] "/Users/mattsettles/Data_in_R"
dir()
## [1] "bwa_mem_Stats.log" "bwa.samtools.stats" "data_in_R_files"
## [4] "data_in_R.html" "data_in_R.knit.md" "data_in_R.md"
## [7] "data_in_R.nb.html" "data_in_R.Rmd" "Data_in_R.Rproj"
## [10] "data_in_R.utf8.md" "indel_ratio.pdf" "indel_ratio.png"
## [13] "insert_size.png" "multi_plot.pdf" "multi_plot.png"
## [16] "packrat"
file.exists("bwa.samtools.stats")
## [1] TRUE
If it returned TRUE, great! If not return to the Prepare data_in_R doc and follow the directions to get the file.
So lets read in the file and view the first few lines and get the length
data <- readLines("bwa.samtools.stats")
head(data)
## [1] "# This file was produced by samtools stats (1.9+htslib-1.9) and can be plotted using plot-bamstats"
## [2] "# This file contains statistics for all reads."
## [3] "# The command line was: stats -@ 32 bwa.bam"
## [4] "# CHK, Checksum\t[2]Read Names\t[3]Sequences\t[4]Qualities"
## [5] "# CHK, CRC32 of reads which passed filtering followed by addition (32bit overflow)"
## [6] "CHK\t77e64415\t5b47b901\t532fe148"
tail(data)
## [1] "GCD\t19.0\t58.824\t0.007\t0.007\t0.007\t0.007\t0.007"
## [2] "GCD\t36.0\t70.588\t0.007\t0.007\t0.007\t0.007\t0.007"
## [3] "GCD\t38.0\t76.471\t0.007\t0.007\t0.007\t0.007\t0.007"
## [4] "GCD\t41.0\t82.353\t0.007\t0.007\t0.007\t0.007\t0.007"
## [5] "GCD\t42.0\t88.235\t0.007\t0.007\t0.007\t0.007\t0.007"
## [6] "GCD\t48.0\t100.000\t0.007\t0.007\t0.007\t0.007\t0.007"
length(data)
## [1] 1866
There are many sections to the samtools stats output, each section begins with a two or three letter code.
With the exception of Summary Numbers, most sections are tables of data, the file explains the format of the data tables, open the log file (in Rstudio is fine) and search for the term ‘grep’.
Lets take a quick look at the comments in the file
grep("^# ",data, value=TRUE)
First lets extract the Summary numbers and create a summary table
?separate
sn <- grep("^SN",data, value=TRUE)
sn <- separate(data.frame(sn),col=1, into=c("ID", "Name","Value"), sep="\t")[,-1]
kable(sn, caption="Summary numbers")
Name | Value |
---|---|
raw total sequences: | 913311962 |
filtered sequences: | 0 |
sequences: | 913311962 |
is sorted: | 0 |
1st fragments: | 456655981 |
last fragments: | 456655981 |
reads mapped: | 800365919 |
reads mapped and paired: | 748856756 |
reads unmapped: | 112946043 |
reads properly paired: | 306860552 |
reads paired: | 913311962 |
reads duplicated: | 0 |
reads MQ0: | 439677889 |
reads QC failed: | 0 |
non-primary alignments: | 290462657 |
total length: | 127407018699 |
total first fragment length: | 58451965568 |
total last fragment length: | 68955053131 |
bases mapped: | 111789284981 |
bases mapped (cigar): | 53892754351 |
bases trimmed: | 0 |
bases duplicated: | 0 |
mismatches: | 1041917776 |
error rate: | 1.933317e-02 |
average length: | 139 |
average first fragment length: | 128 |
average last fragment length: | 151 |
maximum length: | 151 |
maximum first fragment length: | 128 |
maximum last fragment length: | 151 |
average quality: | 26.6 |
insert size average: | 176.9 |
insert size standard deviation: | 132.5 |
inward oriented pairs: | 122015428 |
outward oriented pairs: | 32504015 |
pairs with other orientation: | 4311328 |
pairs on different chromosomes: | 215597607 |
percentage of properly paired reads (%): | 33.6 |
?kable
On your own: While the Value column is numeric, by default it is being read in as characters. Lets use kable align parameter to left justify name and right justify value.
First lets extract the read length data and create a table
rl <- grep("^RL",data, value=TRUE)
rl <- separate(data.frame(rl),col=1, into=c("ID", "read_length", "count"), sep="\t", convert = TRUE)[,-1]
First lets extract the insert sizes data and create a table
is <- grep("^IS",data, value=TRUE)
is <- separate(data.frame(is),col=1, into=c("ID", "insert size","all pairs", "inward", "outward", "other"), sep="\t", convert=TRUE)[,-1]
First lets extract the base composition of first and last pairs and create a table
actg <- grep("^GCC",data, value=TRUE)
actg <- separate(data.frame(actg),col=1, into=c("ID", "cycle", "A", "C", "G", "T", "N", "O"), sep="\t", convert=TRUE)[,-1]
First lets extract the fragment qualities of first and last pairs and create a table
fq <- grep("^FFQ|^LFQ",data, value=TRUE)
fq <- separate(data.frame(fq),col=1, into=c("Pair", "Cycle", seq(41)), sep="\t", convert=TRUE)
## Warning: Expected 43 pieces. Missing pieces filled with `NA` in 279
## rows [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
## 20, ...].
We get a message here, saying data is missing. This is because there are no 38,39,40,41 quality scores (the typical range for Illumina qualities).
First lets extract the GC content of first and last pairs and create a table
gc <- grep("^GCF|^GCL",data, value=TRUE)
gc <- separate(data.frame(gc),col=1, into=c("Pair", "GC", "Count"), sep="\t", convert=TRUE)
First lets extract the indel distribution data and create a table
id <- grep("^ID",data, value=TRUE)
id <- separate(data.frame(id),col=1, into=c("ID", "length", "insertion_count", "deletion_count"), sep="\t", covert=TRUE)[,-1]
First lets extract the indel by cycle data and create a table
ic <- grep("^IC",data, value=TRUE)
ic <- separate(data.frame(ic),col=1, into=c("ID", "cycle", "ins_fwd", "ins_rev", "del_fwd", "del_rev"), sep="\t", convert=TRUE)[,-1]
On your own: Use what you learned above to extract these 2 sections from the file.
Coverage data * First extract the right rows, these begin (^) with IS. * Then turn it into a table using the function separate (View the help of separate) * with 6 columns (ID, coverage_range, coverage, bases) * separate by the tab character “” * and remove the first column ‘[,-1]’, the COV
GC Coverage data * First extract the right rows, these begin (^) with GCD. * Then turn it into a table using the function separate (View the help of separate) * with 8 columns (ID, GC, GC_percentile, gc.10, gc.25, gc.50, gc.75, gc.90) * separate by the tab character “” * and remove the first column ‘[,-1]’, the GCD
summarize(is,low=min(`insert size`), max=max(`insert size`), average=mean(`all pairs`), noutward=sum(outward), ninward=sum(inward))
## low max average noutward ninward
## 1 0 576 272522 32273430 120770240
new_is <- mutate(is,poutward=outward/`all pairs`, pinward=inward/`all pairs`)
On your own Tasks
Try using “distinct”, on is (or new_is) on the outward and inward columns
So now we have new objects (data.frames) that hold the data we are interested in plotting
ggplot2 uses a basic syntax framework (called a Grammar in ggplot2) for all plot types:
A basic ggplot2 plot consists of the following components:
The basic idea: independently specify plot building blocks and combine them (using ‘+’) to create just about any kind of graphical display you want.
ggplot (data = ) +
We use the ggplot function and define the data as ‘is’ and x, y as as.numeric(get(“insert size”)), as.numeric(get(“all pairs”)), respectively. We use “get” because they have spaces in the names, and as.numeric because the data are characters (due to the manner in which we readin the data.
g <- ggplot(data = is)
g + geom_line( aes(x=get("insert size"), y=get("all pairs")))
Ok, now lets add some labels to the plot
g + geom_line( aes(x=get("insert size"), y=get("all pairs"))) +
labs( x = "insert size", y = "all pairs", title ="Mapped insert sizes", subtitle = "All Pairs", caption = "all pairs insert size")
Ok, what about plotting multiple data objects on the same plot (multiple lines), in that case we can specifically set the y axis in geom_line and color, then call geom_lines twice (or more times).
g <- ggplot(data = is, aes(x=get("insert size")))
g + geom_line(aes(y=get("inward")),color="blue") +
geom_line(aes(y=get("outward")),color="orange") +
labs( x = "insert size", y = "all pairs", title ="Mapped insert sizes", subtitle = "All Pairs", caption = "all pairs insert size")
lets try adjusting the x/y limits to 0,600 and 0,20000 respectively.
g + geom_line(aes(y=get("inward")),color="blue") +
geom_line(aes(y=get("outward")),color="orange") +
coord_cartesian(xlim=c(0,500), ylim=c(0,600000))
Ok so now put all these elements together into a single plot, save final plot as ‘g’
** On your own**: Ok so put it all together, plot all four columns of the insert size data object, add in legends, reasonable coordinate limits
** On your own**: Play with ggplot2 themes (ex. theme_classic() )
g <- ggplot(data = is, aes(x=get("insert size")))
g <- g + geom_line(aes(y=get("all pairs")), color="black") +
geom_line(aes(y=get("inward")),color="blue") +
geom_line(aes(y=get("outward")),color="orange") +
geom_line(aes(y=get("other")), color="green")
g <- g +
labs( x = "insert size", y = "all pairs", title ="Mapped insert sizes", subtitle = "All Pairs", caption = "all pairs insert size")
g <- g + coord_cartesian(xlim=c(0,500), ylim=c(0,600000))
g <- g + theme_light()
plot(g)
In order to plot GC percentage we first need to convert the counts to proportions, to do so we can divide the counts by the sum of counts.
head(gc)
## Pair GC Count
## 1 GCF 0.25 9986
## 2 GCF 1.01 6442
## 3 GCF 1.76 6816
## 4 GCF 2.51 8029
## 5 GCF 3.27 9586
## 6 GCF 4.02 11557
h <- ggplot(gc, aes(as.numeric(GC), Count/sum(Count),color=Pair))
h <- h + geom_line()
h
** On your own**: Finish the plot (add labels, etc.). Save the final graph object in h
Sometimes we may need to transform our data before plotting. The melt funciton from reshape2 takes data in wide format (data are in columns) and stacks a set of columns into a single column of data. In the ACTG object we can stack bases values by cycle.
actgm <- melt(actg,id="cycle")
now head the new actgm object. What did melt do?
ic <- ggplot(actgm, aes(as.numeric(cycle), as.numeric(value), by=variable, colour=variable))
i <- ic + geom_line() + coord_cartesian(ylim=c(0,100))
i
** On your own**: Using what you learned until now, finish the plot, save it as object i
i2 <- ic + geom_boxplot()
i2
** On your own**: Try some other geometries (Ex. bin2d, col, count, which generate an ‘interpretable’ plot)
First lets melt the quality scores
fqm <- melt(fq,id=c("Pair","Cycle"))
Take a look at the new object
j <- ggplot(fqm, aes(Cycle, variable))
j + geom_tile(aes(fill = as.numeric(value)))
Now lets try changing the gradient colors and modify the legend, add labels. The ggplot2 ‘theme’ function can be used to modify individual components of a theme.
?theme
j = j + geom_tile(aes(fill = as.numeric(value))) +
scale_fill_gradient(low = "red", high = "green") +
ylab("Cycle") +
xlab("Quality") +
theme(legend.title = element_text(size = 10),
legend.text = element_text(size = 12),
plot.title = element_text(size=16),
axis.title=element_text(size=14,face="bold"),
axis.text.x = element_text(angle = 90, hjust = 1)) +
labs(fill = "Quality value")
j
** On your own** Try modifying scale_fill_gradient to scale_fill_distiller.
** On your own** Play with parts of the plotting function, see how the change modifies the plot.
** On your own** Recreate the indel lengths plot
k <- ggplot(id, aes(x=as.numeric(length)))
k <- k + geom_line(aes(y=as.numeric(insertion_count)), color = "red", size=1.5)
k <- k + geom_line(aes(y=as.numeric(deletion_count)), color = "black", size=1.5)
k
Lets try changing the Y axis to log scale
k <- k + scale_y_log10()
k
Tweek the grid elments using theme
k <- k + theme(panel.grid.minor = element_blank(),
panel.grid.major = element_line(color = "gray50", size = 0.5),
panel.grid.major.x = element_blank())
k
## Warning: Transformation introduced infinite values in continuous y-axis
k <- k + xlab("indel length") + ylab("indel count (log10)")
k
## Warning: Transformation introduced infinite values in continuous y-axis
Now lets also plot the ratio of the 2, but first we need to create the object
id$ratio <- as.numeric(id$insertion_count)/as.numeric(id$deletion_count)
l <- ggplot(id, aes(x=as.numeric(length)))
l <- l + geom_line(aes(y=as.numeric(ratio)), color = "green", size=1.0)
l
Tweek the grid
l <- l + theme(panel.grid.minor = element_blank(),
panel.grid.major = element_line(color = "gray50", size = 0.5),
panel.grid.major.x = element_blank())
l
Update axis labels
l <- l + xlab("indel length") + ylab("insertion/deletion ratio")
l
Now lets use gridExtra to plot both in the same plat
grid.arrange(k, l, nrow = 1)
## Warning: Transformation introduced infinite values in continuous y-axis
The gridExtra package is great for plotting multiple object in one plot.
include_graphics("https://raw.githubusercontent.com/ucdavis-bioinformatics-training/2018-September-Bioinformatics-Prerequisites/master/thursday/Data_in_R/grid_plot.png")
full <- grid.arrange(
g, h, i, i2,
widths = c(2, 1, 1),
layout_matrix = rbind(c(1, 2, NA),
c(3, 3, 4))
)
** on your own**: Play with th grid.arrange function, using the plots you’ve created to create you own final combined plot.
This must be done outside of the Notebook as the notebook expects you to plot in the notebook only, so run on the Console.
Saving plots to pdf ** do on the console **
ggsave("multi_plot.pdf",full,device="pdf",width=6,height=4, units="in", dpi=300)
Saving plots to png ** do on the console **
ggsave("multi_plot.png",full,device="png",width=6,height=4, units="in", dpi=300)
View the help documentation for ggsave, what other
With any remaining time (or homework), use the ggplot cheat sheet to further expand and modify the plots.