how can I turn a skinny/long data frame into a Seurat object

Question

[glocke01]

I'm trying to turn a long formatted tibble/data frame into a Seurat object. So, my input has one row per cell per gene:

library(tibble)
library(Seurat)
n_rows <- 10000
cnt_frame <- tibble(cell_id = sprintf("sample%02d_cell%03d", 
                                      sample(1:10, n_rows, replace=T), 
                                      sample(1:1000, n_rows, replace=T)),
                    gene_id = sprintf("TLA%03d", sample(1:500, n_rows, replace=T)),
                    read_count = rnbinom(n_rows, size=10, mu=10))
head(cnt_frame)
# A tibble: 6 x 3
  cell_id          gene_id read_count
  <chr>            <chr>   <dbl>
1 sample08_cell813 TLA092      6
2 sample10_cell481 TLA167     10
3 sample03_cell173 TLA029     14
4 sample07_cell140 TLA358     10
5 sample03_cell021 TLA314      9
6 sample08_cell091 TLA228      8

I now wish to turn this into a Seurat object.

cnt_mat <- cnt_frame %>%
  pivot_wider(names_from=cell_id, values_from=read_count, values_fill=0) %>% ## how to do without pivot?
  column_to_rownames("gene_id") %>%
  Seurat::as.sparse()
seur <- CreateSeuratObject(cntMat, meta.data=my_metadata)

My question is how to do this without the pivot_wider. My actual data frame is much bigger than this toy example and the pivot step is reaching integer overflow. Given that a sparse matrix is, in principle, quite close to a long table, pivoting to a full matrix seems like it should be unnecessary, but I can't figure out how to get from here to there.

Answer 1

[rpolicastro]

Slightly modified example data.

cnt_mat <- structure(list(cell_id = c("sample08_cell813", "sample10_cell481",
"sample03_cell173", "sample07_cell140", "sample03_cell021", "sample08_cell091",
"sample08_cell091"), gene_id = c("TLA092", "TLA167", "TLA029",
"TLA358", "TLA314", "TLA228", "TLA029"), read_count = c(6, 10,
14, 10, 9, 8, 2)), row.names = c(NA, -7L), class = "data.frame")

> cnt_mat
           cell_id gene_id read_count
1 sample08_cell813  TLA092          6
2 sample10_cell481  TLA167         10
3 sample03_cell173  TLA029         14
4 sample07_cell140  TLA358         10
5 sample03_cell021  TLA314          9
6 sample08_cell091  TLA228          8
7 sample08_cell091  TLA029          2

You can construct the sparse matrix directly. This is assuming that there are no 0 counts in the long data (might want to double check to be sure).

library("Matrix")

sparse_mat <- sparseMatrix(
  i=match(cnt_mat$gene_id, unique(cnt_mat$gene_id)),
  j=match(cnt_mat$cell_id, unique(cnt_mat$cell_id)),
  x=cnt_mat$read_count,
  dims=c(length(unique(cnt_mat$gene_id)), length(unique(cnt_mat$cell_id))),
  dimnames=list(unique(cnt_mat$gene_id), unique(cnt_mat$cell_id))
)

> sparse_mat
6 x 6 sparse Matrix of class "dgCMatrix"
       sample08_cell813 sample10_cell481 sample03_cell173 sample07_cell140
TLA092                6                .                .                .
TLA167                .               10                .                .
TLA029                .                .               14                .
TLA358                .                .                .               10
TLA314                .                .                .                .
TLA228                .                .                .                .
       sample03_cell021 sample08_cell091
TLA092                .                .
TLA167                .                .
TLA029                .                2
TLA358                .                .
TLA314                9                .
TLA228                .                8

[glocke01]

thanks. CreateSeuratObject(sparse_mat) seems to be working as expected. I'll report back once I try implementing with my non-toy data.

[glocke01]

This is terribly obvious in retrospect. The effort of not slapping my forehead is good practice in self-compassion.

[rpolicastro]

No worries! It's not immediately obvious how to generate a sparse matrix by hand like this, so don't beat yourself up.

[glocke01]

yeah, actually, I'm not sure I've ever used match before.