What is the best way to confirm if a mutated gene is more frequent in recurrence group than in baseline?

Question

[Lila M]

Good morning,

I have a data set for different patients and mutated genes in two conditions (recurrence and T0). What I want to compare is if the genes are more frequent at recurrence or T0. I am not considering expression, type of mutation etc. Eg:

# Example data frame
gene_counts <- data.frame(
  gene = c("A", "B", "C", "N", "T", "X", "Y", "Z"),
  Recurrence = c(10, 5, 0, 100, 1, 0, 1, 0),
  T0 = c(50, 10, 4, 150, 0, 1, 0, 1)
)

Now, my question for asses this comparison is, should I check for frequencies to define the type of test (F-fisher or Chi) basedon the frequencies as follow:

# Function to perform the appropriate test based on counts
perform_test <- function(recurrence, t0) {
  contingency_table <- matrix(c(recurrence, t0, sum(gene_counts$Recurrence) - recurrence, sum(gene_counts$T0) - t0), nrow = 2)

  if (any(contingency_table < 5)) {
    # Use Fisher's test if counts are low
    p.value <- fisher.test(contingency_table)$p.value
  } else {
    # Use Chi-square test otherwise
    p.value <- chisq.test(contingency_table)$p.value
  }

  return(p.value)
}

# Apply test for each gene
gene_counts <- gene_counts %>%
  rowwise() %>%
  mutate(p_value = perform_test(Recurrence, T0)) %>%
  ungroup()
gene_counts <- gene_counts %>%
  mutate(adj_p_value = p.adjust(p_value, method = "BH"))

Or is there any other more straight forward way to do this? e.g is there a good practice or standardised statistical test for this approach rather than to check every case?

P.S Not all the genes are necessarily present in all the samples

Thank you!!!

[Pierre Lindenbaum]

rvtest ? http://zhanxw.github.io/rvtests/

skat ? skat-o ? https://cran.r-project.org/web/packages/SKAT/SKAT.pdf

regenie ? https://rgcgithub.github.io/regenie/

[Lila M]

Those are really good tools, however, I am not sure if suitable for my approach? As what I have is the counts per gene per condition within a population, not directly related with specific mutations. Would a permutation test be beneficial in this case?

gene Recurrence  T0
1    A         10  50
2    B          5  10
3    C          0   4
4    N        100 150
5    T          1   0

[Pierre Lindenbaum]

As what I have is the counts per gene per condition within a population,

yes, I'm not a specialist, but I think the tools above don't use the specific variations, just looking how many people in case/control population have least one zero rare allele (missense...) in a gene.

[Lila M]

Thank you, The idea for this is just to compare if there are differences in the frequency of a mutated gene in a population at two different time points. I am struggle try to find the best model to do this comparison... Hopefully someone here might have more experience!