Convert a list of character vectors into a binary matrix

user_16240593 0
Viewed 155

I am working in R and have a named list of character vectors. Each vector describes the genes present in a biological pathway.

Please see example below:

gene_sets = list(pathwayX= c("Gene3"),pathwayY= c("Gene2", "Gene3"),pathwayz= c("Gene1", "Gene2","Gene3"))

> gene_sets
$pathwayX
"Gene3"

$pathwayY
"Gene2" "Gene3"

$pathwayZ
"Gene1" "Gene2" "Gene3"

What I want to do is to convert the list into a binary matrix, with genes as columns and pathways as rows. 0 means gene is not present in the pathway and 1 means gene is present.

Gene1 Gene2 Gene3
pathwayX 0 0 1
pathwayY 0 1 1
pathwayZ 1 1 1
vector r matrix list
5 Answers

We can use

t(+sapply(gene_sets, "%in%", x = c("Gene1", "Gene2", "Gene3")))

If you want to dynamically obtain c("Gene1", "Gene2", "Gene3"), we can do

GeneID <- sort(unique(unlist(gene_sets)))

mat <- t(+sapply(gene_sets, "%in%", x = GeneID))  ## matrix output
colnames(mat) <- GeneID
#         Gene1 Gene2 Gene3
#pathwayX     0     0     1
#pathwayY     0     1     1
#pathwayz     1     1     1

data.frame(mat)  ## data.frame output

My impression is that genes problem is usually large and sparse. If you have hundreds of thousands of genes and pathways in reality, the following sparse matrix solution is the best option.

pathwayID <- names(gene_sets)
n1 <- lengths(gene_sets, use.names = FALSE)  ## number of genes in each pathway
genesVec <- unlist(gene_sets, use.names = FALSE)
GeneID <- sort(unique(genesVec))
i <- rep(1:length(n1), n1)
j <- match(genesVec, GeneID)
Matrix::sparseMatrix(i = i, j = j, x = rep.int(1, length(i)),
                     dimnames = list(pathwayID, GeneID))
#3 x 3 sparse Matrix of class "dgCMatrix"
#         Gene1 Gene2 Gene3
#pathwayX     .     .     1
#pathwayY     .     1     1
#pathwayz     1     1     1
user_48917380

An idea is to use table in combination with stack i.e.

t(table(stack(gene_sets)))

          values
ind        Gene1 Gene2 Gene3
  pathwayX     0     0     1
  pathwayY     0     1     1
  pathwayz     1     1     1
user_56355800

You could also use mtabulate from qdapTools like this:

gene_sets = list(pathwayX= c("gene3"),pathwayY= c("gene2", "gene3"),pathwayz= c("gene1", "gene2","gene3"))

library(qdapTools)
mtabulate(gene_sets)
#>          gene1 gene2 gene3
#> pathwayX     0     0     1
#> pathwayY     0     1     1
#> pathwayz     1     1     1

Created on 2022-07-18 by the reprex package (v2.0.1)

user_142827140

Here is a try to solve it with tidyverse:

library(dplyr)
library(tibble)
as.data.frame(unlist(gene_sets)) %>% 
  transmute(gene = as.factor(`unlist(gene_sets)`)) %>% 
  rownames_to_column() %>% 
  mutate(rowname = str_remove(rowname, "[0-9]")) %>% 
  cbind((model.matrix(~ gene + 0, .) == 1)*1) %>% 
  rename_with(., ~str_replace_all(., "geneGene", "Gene")) %>% 
  group_by(rowname) %>% 
  summarise(across(-gene, ~sum(.)))

  rowname  Gene1 Gene2 Gene3
  <chr>    <dbl> <dbl> <dbl>
1 pathwayX     0     0     1
2 pathwayY     0     1     1
3 pathwayz     1     1     1
user_133216470

Here is a somewhat concise tidyverse way if one were compelled to go down that path.

library(dplyr)
library(purrr)
library(tibble)
library(tidyr)

map_dfr(gene_sets,
        ~ as_tibble_row(set_names(rep(1L, length(.x)), .x)), .id = "row") %>% 
  relocate(row, order(colnames(.))) %>% 
  mutate(across(-1, replace_na, 0))

# A tibble: 3 x 4
  row      Gene1 Gene2 Gene3
  <chr>    <int> <int> <int>
1 pathwayX     0     0     1
2 pathwayY     0     1     1
3 pathwayz     1     1     1
user_109174790
Related