Replace values in R data frame according to values in other data frame

Question

I would like to change row names values from a first data frame if these rows names are present in another data frame, and change it to a corresponding value (defined in the second data frame). I could do it with a for loop, but I wonder what would be some other more suitable ways to do it in R.

Here is a toy dataset (my real dataset has 10^5 rows):

df <- data.frame("sample_1" = rep(0,3), "sample_2" = rep(0,3), row.names = paste0('gene_', seq(1, 3)), stringsAsFactors = FALSE)
annotation <- data.frame("gene" =  paste0('gene_', seq(1, 2)), 'name' = paste0('name_', seq(1, 2)), stringsAsFactors = FALSE)
df$gene <- rownames(df)

Initial data frames:

df 
       sample_1 sample_2   gene
gene_1        0        0 gene_1
gene_2        0        0 gene_2
gene_3        0        0 gene_3

annotation
    gene   name
1 gene_1 name_1
2 gene_2 name_2

Solution found:

for (x in rownames(df)){
  if (x %in% annotation$gene){
    df[x,]$gene <- annotation$name[which(annotation$gene == x)]
  }
}
rownames(df) <- df$gene
df$gene <- NULL

Resulting data frame:

df
       sample_1 sample_2
name_1        0        0
name_2        0        0
gene_3        0        0

Answer 1

Code Review

One change I would make to your solution is to directly modify the row names as below (instead of modifying gene and then setting the row names equal to gene).

for (i in 1:nrow(df)){
  if (rownames(df)[i] %in% annotation$gene) {

    rownames(df)[i] = annotation$name[which(annotation$gene == rownames(df)[i])]

  }
}
df$gene <- NULL

df
#        sample_1 sample_2
# name_1        0        0
# name_2        0        0
# gene_3        0        0

This solution is similar to the original, but iterates over row numbers (1-3 in this case) instead of row names. This allows direct modification of the row names and makes the code slightly more concise. This version doesn't use gene at all.

That's how I'd modify the existing solution, but overall I wouldn't use a for loop. for loops can be slow since they iterate over every element (and your data set has 10^5 rows). Below is another approach using the dplyr library.

Another Approach

library(dplyr)
library(tibble)

df <- left_join(df, annotation, by = "gene") %>%        # Join "annotation" and "df"
mutate(gene = if_else(is.na(name), gene, name)) %>%     # Convert "gene" to "name" when "name" is valid
column_to_rownames(var = "gene") %>%                    # Set row names to "gene"
select(-name)                                           # Remove "name"

df
#        sample_1 sample_2
# name_1        0        0
# name_2        0        0
# gene_3        0        0

This syntax may look unfamiliar but it's doing 3 main things:
- left-joining df and annotation by the gene field. It has to be a left-join so all rows of df are preserved!
- changing the value of gene to name only when name is non-null
- setting the row names equal to gene

The pipe operator %>% passes the data through these operations without needing to repeatedly type df <-. These are all vectorized operations and so are faster than a for loop, especially for large data sets.

dplyr syntax can be intimidating at first, but it's an extremely popular (and I believe intuitive) paradigm to manipulate data. Here is a handy intro on dplyr syntax.