2

I have a data frame like following (the rownames are "1", "2", "3"...). Since there are non unique entries in each column, I cannot assign any of them as row names.

gene cell count
a    c1    1
a    c2    1
a    c3    4
b    c1    3
b    c2    1
b    c3    1
f    c1    3
d    c8    9
e    c11   1

Each gene is measured in each cell (means they have a value in count column) but zero counts are not shown (for example gene "a" has zero counts in cells c8 and c11, hence do not appear).

Now I want to reshape/convert the data frame into dgCMatrix with following arrangement

(genes as row names, cells as column names and count values as data points)

   c1  c2  c3  c8  c11 
a  1   1   4   .    .
c  3   1   1   .    .

where "." corresponds to a zero count.

I tried reshape, reshape2, as.matrix as mentioned in many posts here, but no success.

James Z
  • 12,209
  • 10
  • 24
  • 44
Tushar Deshpande
  • 23
  • 1
  • 3

1 Answers1

2

You convert to long format and set the gene column as rownames first:

library(Matrix)
library(dplyr)
library(tidyr)

mat <- df %>% pivot_wider(id_cols = gene,values_from = count,names_from = cell,
values_fill = list(count=0)) %>% tibble::column_to_rownames("gene")

Then to sparseMatrix:

mat = Matrix(as.matrix(mat),sparse=TRUE)

    5 x 5 sparse Matrix of class "dgCMatrix"
  c1 c2 c3 c8 c11
a  1  1  4  .   .
b  3  1  1  .   .
f  3  .  .  .   .
d  .  .  .  9   .
e  .  .  .  .   1
StupidWolf
  • 45,075
  • 17
  • 40
  • 72