Create a n x n matrix of counts using R-CodePudding

I have a boolean matrix with observations in row × features in columns.

> df
      TP53 PIK3CA GATA3 MAP3K1  PTEN KMT2C  AKT1   NF1 FOXA1 SF3B1
ACYZ FALSE  FALSE FALSE  FALSE FALSE FALSE FALSE FALSE FALSE FALSE
CKPD FALSE   TRUE FALSE  FALSE FALSE FALSE FALSE FALSE FALSE FALSE
AEXF FALSE  FALSE FALSE  FALSE FALSE FALSE  TRUE FALSE FALSE FALSE
AGNS FALSE  FALSE FALSE  FALSE FALSE FALSE FALSE FALSE FALSE FALSE
AJEH  TRUE  FALSE FALSE  FALSE FALSE FALSE FALSE FALSE FALSE FALSE
AJYK  TRUE   TRUE FALSE  FALSE FALSE FALSE  TRUE FALSE FALSE FALSE
AKVJ  TRUE  FALSE FALSE  FALSE FALSE FALSE FALSE  TRUE FALSE FALSE
ALGN FALSE  FALSE FALSE   TRUE  TRUE  TRUE FALSE FALSE FALSE FALSE
AMFQ FALSE   TRUE FALSE  FALSE FALSE FALSE FALSE FALSE FALSE FALSE
ANPD FALSE  FALSE FALSE  FALSE FALSE  TRUE FALSE FALSE FALSE FALSE

I want to create a symmetric matrix (features × features) based on the matrix above and the values in the cells are the counts of observations (not binary) that meet the features in both rows and columns (TRUE-TRUE).

> output
         TP53 PIK3CA GATA3 MAP3K1  PTEN KMT2C  AKT1   NF1 FOXA1 SF3B1
TP53        0      1     0      0     0     0     1     1     0     0
PIK3CA      1      0     0      0     0     0     1     0     0     0
...

data

structure(c(FALSE, FALSE, FALSE, FALSE, TRUE, TRUE, TRUE, FALSE, 
FALSE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, TRUE, FALSE, 
FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, 
FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, 
FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, 
FALSE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE, 
FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, TRUE, FALSE, FALSE, 
TRUE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE, 
FALSE, FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, 
FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, 
FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, 
FALSE, FALSE), .Dim = c(10L, 10L), .Dimnames = list(c("ACYZ", 
"CKPD", "AEXF", "AGNS", "AJEH", "AJYK", "AKVJ", "ALGN", "AMFQ", 
"ANPD"), c("TP53", "PIK3CA", "GATA3", "MAP3K1", "PTEN", "KMT2C", 
"AKT1", "NF1", "FOXA1", "SF3B1")))

Thanks.

CodePudding user response：

This is a version in which cells (row, col) = (CKPD, PIK3CA) and (PIK3CA, CKPD) are treated differently resulting in output cells containing either 0 or 1.

library(tidyverse)

data <- structure(c(
  FALSE, FALSE, FALSE, FALSE, TRUE, TRUE, TRUE, FALSE,
  FALSE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, TRUE, FALSE,
  FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE,
  FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE,
  FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE,
  FALSE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE,
  FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, TRUE, FALSE, FALSE,
  TRUE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE, FALSE, FALSE,
  FALSE, FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, FALSE, FALSE,
  FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE,
  FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE, FALSE,
  FALSE, FALSE
), .Dim = c(10L, 10L), .Dimnames = list(c(
  "ACYZ",
  "CKPD", "AEXF", "AGNS", "AJEH", "AJYK", "AKVJ", "ALGN", "AMFQ",
  "ANPD"
), c(
  "TP53", "PIK3CA", "GATA3", "MAP3K1", "PTEN", "KMT2C",
  "AKT1", "NF1", "FOXA1", "SF3B1"
)))

all_features <-
  data %>%
  rownames() %>%
  union(colnames(data))

data %>%
  as_tibble(rownames = "from") %>%
  pivot_longer(-from, names_to = "to") %>%
  group_by(from, to) %>%
  mutate(
    from = from %>% factor(levels = all_features),
    to = to %>% factor(levels = all_features),
    value = sum(value)
  ) %>%
  ungroup() %>%
  complete(from, to, fill = list(value = 0)) %>%
  pivot_wider(names_from = to, values_from = value)
#> # A tibble: 20 × 21
#>    from    ACYZ  CKPD  AEXF  AGNS  AJEH  AJYK  AKVJ  ALGN  AMFQ  ANPD  TP53
#>    <fct>  <int> <int> <int> <int> <int> <int> <int> <int> <int> <int> <int>
#>  1 ACYZ       0     0     0     0     0     0     0     0     0     0     0
#>  2 CKPD       0     0     0     0     0     0     0     0     0     0     0
#>  3 AEXF       0     0     0     0     0     0     0     0     0     0     0
#>  4 AGNS       0     0     0     0     0     0     0     0     0     0     0
#>  5 AJEH       0     0     0     0     0     0     0     0     0     0     1
#>  6 AJYK       0     0     0     0     0     0     0     0     0     0     1
#>  7 AKVJ       0     0     0     0     0     0     0     0     0     0     1
#>  8 ALGN       0     0     0     0     0     0     0     0     0     0     0
#>  9 AMFQ       0     0     0     0     0     0     0     0     0     0     0
#> 10 ANPD       0     0     0     0     0     0     0     0     0     0     0
#> 11 TP53       0     0     0     0     0     0     0     0     0     0     0
#> 12 PIK3CA     0     0     0     0     0     0     0     0     0     0     0
#> 13 GATA3      0     0     0     0     0     0     0     0     0     0     0
#> 14 MAP3K1     0     0     0     0     0     0     0     0     0     0     0
#> 15 PTEN       0     0     0     0     0     0     0     0     0     0     0
#> 16 KMT2C      0     0     0     0     0     0     0     0     0     0     0
#> 17 AKT1       0     0     0     0     0     0     0     0     0     0     0
#> 18 NF1        0     0     0     0     0     0     0     0     0     0     0
#> 19 FOXA1      0     0     0     0     0     0     0     0     0     0     0
#> 20 SF3B1      0     0     0     0     0     0     0     0     0     0     0
#> # … with 9 more variables: PIK3CA <int>, GATA3 <int>, MAP3K1 <int>, PTEN <int>,
#> #   KMT2C <int>, AKT1 <int>, NF1 <int>, FOXA1 <int>, SF3B1 <int>

^{Created on 2022-03-14 by the reprex package (v2.0.0)}

Note

If you want to do co-occurrence (e.g. samples containing both genes) or correlation with numerical outcomes, you want to sum up counts from the cell (a,b) and (b,a). This will result in cells having either 0, 1, or 2. This is not what this answer is about, but the mathematical word symmetric matrix would require this to do so.