calculate reads per million mapped read using R - r

Question

df1 <- read.table(text="
gene_id A1 A2 A3 A4 length Total
ENSMUSG00000000028 58 93 48 58 789 200
ENSMUSG00000000031 11 7 20 16 364 54
ENSMUSG00000000037 3 5 6 98 196 112
ENSMUSG00000000058 66 93 69 71 436 299
ENSMUSG00000000085 55 68 97 67 177 287", header=TRUE)
The table represents the read count in a gene in different samples (A1, A2..A4).
How can i calculate the reads per million mapped read (RPKM) for these raw read counts using R
RPKM = (number of reads in a gene * 1e6)/(Total*length)
out_put <- read.table(text="
gene_id A1 A2 A3 A4
ENSMUSG00000000028 367.5539 589.3536 304.1825 367.5539
ENSMUSG00000000031 559.6256 356.1254 1017.5010 814.0008
ENSMUSG00000000037 136.6618 227.7697 273.3236 4464.2857
ENSMUSG00000000058 506.2747 713.3871 529.2872 544.6289
ENSMUSG00000000085 1082.6985 1338.6090 1909.4864 1318.9236", header=TRUE)

Answer 1

One way to do this without writing lines or a loop is using melt and dcast:
library(reshape2)
m_df1 <- melt(df1, measure.vars=c("A1","A2","A3","A4"))
m_df1$RPKM <- with(m_df1, value*1e6 / (Total*length))
output <- dcast(gene_id~variable,value.var="RPKM",data=m_df1)
> output
gene_id A1 A2 A3 A4
1 ENSMUSG00000000028 367.5539 589.3536 304.1825 367.5539
2 ENSMUSG00000000031 559.6256 356.1254 1017.5010 814.0008
3 ENSMUSG00000000037 136.6618 227.7697 273.3236 4464.2857
4 ENSMUSG00000000058 506.2747 713.3871 529.2872 544.6289
5 ENSMUSG00000000085 1082.6985 1338.6090 1909.4864 1318.9236
A second way is to use sapply to create a matrix of estimates, which you can then either rename and add to your original data, or cbind to your gene_ids.
my_cols <- c("A1","A2","A3","A4")
RPKMs <- sapply(my_cols, function(x){
df1[,x]*1e6/(df1$Total*df1$length)
}
)
output <- cbind(df1$gene_id,RPKMs)

Answer 2

You can achieve this also without reshaping. Using the data.table package:
library(data.table)
setDT(df1)[,indx:=.I][, lapply(.SD, function(x) (x * 1e6) / (Total * length)),
by=.(indx,gene_id,length,Total)]
this gives:
indx gene_id length Total A1 A2 A3 A4
1: 1 ENSMUSG00000000028 789 200 367.5539 589.3536 304.1825 367.5539
2: 2 ENSMUSG00000000031 364 54 559.6256 356.1254 1017.5010 814.0008
3: 3 ENSMUSG00000000037 196 112 136.6618 227.7697 273.3236 4464.2857
4: 4 ENSMUSG00000000058 436 299 506.2747 713.3871 529.2872 544.6289
5: 5 ENSMUSG00000000085 177 287 1082.6985 1338.6090 1909.4864 1318.9236
Explanation:
with setDT(df1) you convert the dataframe to a datatable
with [,indx:=.I] you create a unique identifier for each row
with by=.(indx,gene_id,length,Total) you determine the columns by which you want to group the data (these columns will not be transformed), by including the indx you make sure that each row is an unique group
with lapply(.SD, function(x) (x * 1e6) / (Total * length)) you apply the required calculation to each column which is not specified in the by statement
A similar solution with dplyr:
library(dplyr)
func <- function(x,y,z) (x * 1e6) / (y * z)
df1 %>% mutate(indx=seq(1,nrow(.))) %>%
group_by(indx,gene_id,length,Total) %>%
summarise_each(funs(func(.,Total,length)))
wich gives:
indx gene_id length Total A1 A2 A3 A4
(int) (fctr) (int) (int) (dbl) (dbl) (dbl) (dbl)
1 1 ENSMUSG00000000028 789 200 367.5539 589.3536 304.1825 367.5539
2 2 ENSMUSG00000000031 364 54 559.6256 356.1254 1017.5010 814.0008
3 3 ENSMUSG00000000037 196 112 136.6618 227.7697 273.3236 4464.2857
4 4 ENSMUSG00000000058 436 299 506.2747 713.3871 529.2872 544.6289
5 5 ENSMUSG00000000085 177 287 1082.6985 1338.6090 1909.4864 1318.9236