Populate matrix by colname identity - r

Question

I have many samples, each one of which has a corresponding abundance matrix. From these abundance matrices, I would like to create a large matrix that contains abundance information for each sample in rows.
For example, a single abundance matrix would look like:
A B C D
sample1 1 3 4 2
where A, B, C, and D represent colnames, and the abundances are the row values.
I would like to populate my larger matrix, which has as colnames all possible letters (A:Z) and all possible samples (sample1:sampleN) as rows, by matching the colname values.
For ex. :
A B C D E F G .... Z
sample1 1 3 4 2 NA NA NA ....
sample2 NA NA 2 5 7 NA NA ....
sample3 4 NA 6 9 2 NA 2 .....
....
sampleN
Different samples have a varying mix of abundances, in no guaranteed order.
When iteratively adding to this larger matrix, how could I ensure that the correct columns are populated by the right abundance values (ex. column "A" is only filled by values corresponding to abundances of "A" in different samples)? Thanks!

Answer 1

Starting data, changing just a little to highlight differences:
m1 <- as.matrix(read.table(header=TRUE, text="
A B C Z
sample1 1 3 4 2"))
m2 <- as.matrix(read.table(header=TRUE, text="
A B C D E F G
sample2 NA NA 2 5 7 NA NA
sample3 4 NA 6 9 2 NA 2"))
First, we need to make sure both matrices have the same column names:
newcols <- setdiff(colnames(m2), colnames(m1))
m1 <- cbind(m1, matrix(NA, nr=nrow(m1), nc=length(newcols), dimnames=list(NULL, newcols)))
newcols <- setdiff(colnames(m1), colnames(m2))
m2 <- cbind(m2, matrix(NA, nr=nrow(m2), nc=length(newcols), dimnames=list(NULL, newcols)))
m1
# A B C Z D E F G
# sample1 1 3 4 2 NA NA NA NA
m2
# A B C D E F G Z
# sample2 NA NA 2 5 7 NA NA NA
# sample3 4 NA 6 9 2 NA 2 NA
And now we combine them; regular cbind needs the column names to be aligned as well:
rbind(m2, m1[,colnames(m2),drop=FALSE])
# A B C D E F G Z
# sample2 NA NA 2 5 7 NA NA NA
# sample3 4 NA 6 9 2 NA 2 NA
# sample1 1 3 4 NA NA NA NA 2

Answer 2

You should be able to take advantage of matrix indexing, like so:
big[cbind(rownames(abun),colnames(abun))] <- abun
Using this example abundance matrix, and a big matrix to fill:
abun <- matrix(c(1,3,4,2),nrow=1,dimnames=list("sample1",LETTERS[1:4]))
big <- matrix(NA,nrow=5,ncol=26,dimnames=list(paste0("sample",1:5),LETTERS))

Answer 3

Another solution using reduce from purrr package and union_all from dplyr package:
library(purrr)
library(dplyr)
sample_names <- c("sample1","sample2","sample3")
Generating 3 random abundance dataframes:
num1 <- round(runif(runif(1,min = 1, max = 10),min = 1, max = 10))
df1 <- data.frame(t(num1))
colnames(df1) <- sample(LETTERS,length(num1))
num2 <- round(runif(runif(1,min = 1, max = 10),min = 1, max = 10))
df2 <- data.frame(t(num2))
colnames(df2) <- sample(LETTERS,length(num2))
num3 <- round(runif(runif(1,min = 1, max = 10),min = 1, max = 10))
df3 <- data.frame(t(num3))
colnames(df3) <- sample(LETTERS,length(num3))
This is actually the code that does all the magic:
A <- reduce(list(df1,df2,df3),union_all)
col_order <- sort(colnames(A),decreasing = FALSE)
A <- A[,col_order]
rownames(A) <- sample_names
Output:
> A
A C E F O P Q U W Y
sample1 9 NA NA NA 9 NA 5 6 NA NA
sample2 NA NA NA NA 5 4 NA NA 5 NA
sample3 NA 6 5 9 NA NA 3 NA 5 7