Recreating (and maybe improving on) some of the figures generated with plot-bamstats application in R.
First load knitr, tidyverse, reshape2 and gridExtra packages.
library(knitr)
library(tidyverse)
library(reshape2)
library(gridExtra)
This document assumes you have the file ‘bwa.samtools.stats’ in your current working directory, test to make sure it is.
getwd()
## [1] "/Users/jli/Jessie/Research/BioInfo/Courses/2019-March-Bioinformatics-Prerequisites/wednesday/Data_in_R"
dir()
## [1] "bwa_mem_Stats" "bwa_mem_Stats.log"
## [3] "bwa.samtools.stats" "bwa.samtools.stats.plot"
## [5] "data_in_R_files" "data_in_R_prepare.md"
## [7] "data_in_R_prepare.nb.html" "data_in_R_prepare.Rmd"
## [9] "data_in_R.html" "data_in_R.log"
## [11] "data_in_R.md" "data_in_R.nb.html"
## [13] "data_in_R.Rmd" "grid_plot.png"
## [15] "multi_plot.pdf" "multi_plot.png"
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 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’.
Take a quick look at the comments in the file
head(grep("^# ",data, value=TRUE))
## [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] "# Summary Numbers. Use `grep ^SN | cut -f 2-` to extract this part."
First 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
Exercise 1: While the Value column is numeric, by default it is being read in as characters. Use kable align parameter to left justify name and right justify value.
First 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 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 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 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 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 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 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]
Exercise 2: Use what you learned above to extract these 2 sections from the file.
Coverage data
GC Coverage data
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`)
Exercise 3: Try using “distinct”, on is (or new_is) on the outward and inward columns.
Exercise 4:
View the head/tail of some (or even all) of the objects.
Use dim to get an idea of the table dimentions.
Use summary to summarize and produce summary statistics (min, max, means, 1st and 3rd quartile boundaries) of the columns.
Any other summaries?
So now we have new objects (data.frames) that hold the data that 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 layers and combine them (using ‘+’) to create just about any kind of graphical display you want.
ggplot (data = <DATA> ) +
<GEOM_FUNCTION> (mapping = aes( <MAPPINGS> ), stat = <STAT> , position = <POSITION> ) +
<COORDINATE_FUNCTION> +
<FACET_FUNCTION> +
<SCALE_FUNCTION> +
<THEME_FUNCTION>
We use the ggplot function and define the data as ‘is’ and x, y as get(“insert size”), get(“all pairs”), respectively. We use “get” because they have spaces in the names. There is another way to do it. Hint: look at the documentation above for clues.
g <- ggplot(data = is)
g + geom_line( aes(x=get("insert size"), y=get("all pairs")))
Ok, now 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)? We can specifically set the y axis in geom_line and color, then call geom_lines a second time (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")
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’
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)
Exercise 5: Put it all together, plot all four columns of the insert size data object, add in legends, reasonable coordinate limits.
Exercise 6: Play with ggplot2 themes (ex. theme_classic() )
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(GC, Count/sum(Count),color=Pair))
h <- h + geom_line()
h
Exercise 7: 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(cycle, value, by=variable, colour=variable))
i <- ic + geom_line() + coord_cartesian(ylim=c(0,100))
i
Exercise 8: Using what you have learned so far, finish the plot, save it as object i.
i2 <- ic + geom_boxplot()
i2
Exercise 9: Try some other geometries (Ex. bin2d, col, count, which generate an ‘interpretable’ plot).
First 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 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
Exercise 10: Try modifying scale_fill_gradient to scale_fill_distiller.
Exercise 11: Play with parts of the plotting function, see how the change modifies the 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
Try changing the Y axis to log scale
k <- k + scale_y_log10()
k
Tweak 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 also plot the ratio of the 2, but first we need to create the ratio data
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
Tweak 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 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))
)
Exercise 12: Play with th grid.arrange function, using the plots you’ve created to create you own final combined plot.
This might have to be done outside of the Notebook as the notebook may expect you to plot in the notebook only, so run on the Console if you have trouble running in the Notebook.
Saving plots to pdf ** do on the console if having trouble in Notebook **
ggsave("multi_plot.pdf",full,device="pdf",width=6,height=4, units="in", dpi=300)
Saving plots to png ** do on the console if having trouble in Notebook **
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.