PottsC
commented
2 years ago Maybe the full traceback for the error will be helpful:
> cluster0_idea <- iDEA.louis(cluster0_idea)
| | 0%, ETA 00:09
Error in if (!is.na(res$annot_coef[2])) { : argument is of length zero
> traceback()
3: stop(results)
2: pbmclapply(1:num_annot, FUN = function(i) {
res <- object@de[[i]]
Annot <- rep(0, object@num_gene)
Annotind = which(names(object@annotation) == names(object@de)[i])
Annot[object@annotation[[Annotind]]] <- 1
Annot <- Annot - mean(Annot)
Annot <- as.matrix(data.frame(rep(1, object@num_gene), Annot))
LouisMethod <- function(res, A) {
numer <- (1 - res$pip) * res$pip * pnorm(res$beta, mean = 0,
sd = sqrt(res$sigma2_beta * res$sigma2_e)) * pnorm(res$beta,
mean = 0, sd = sqrt(res$sigma2_beta2 * res$sigma2_e))
denom <- res$pip * pnorm(res$beta, mean = 0, sd = sqrt(res$sigma2_beta *
res$sigma2_e)) + (1 - res$pip) * pnorm(res$beta,
mean = 0, sd = sqrt(res$sigma2_beta2 * res$sigma2_e))
sel_index <- which(!is.na(numer/denom^2))
numer <- numer[sel_index]
denom <- denom[sel_index]
A <- A[sel_index, ]
T <- matrix(0, ncol = 2, nrow = 2)
T[1, 1] <- sum(as.numeric((numer/denom^2) * A[, 1] *
A[, 1]))
T[1, 2] <- sum(as.numeric((numer/denom^2) * A[, 1] *
A[, 2]))
T[2, 1] <- sum(as.numeric((numer/denom^2) * A[, 1] *
A[, 2]))
T[2, 2] <- sum(as.numeric((numer/denom^2) * A[, 2] *
A[, 2]))
annot_var <- diag(solve(res$info_mat - T))
return(annot_var)
}
if (!is.na(res$annot_coef[2])) {
try_test <- try(louis_var <- LouisMethod(res, Annot),
silent = T)
if (class(try_test) == "try-error") {
resTemp = NULL
}
else {
resTemp = data.frame(annot_id = object@annot_id[Annotind],
annot_coef = res$annot_coef[2], annot_var = res$annot_var[2],
annot_var_louis = louis_var[2], sigma2_b = res$sigma2_beta)
}
}
else {
resTemp = NULL
}
}, mc.cores = getOption("mc.cores", num_core))
1: iDEA.louis(cluster0_idea)
YingMa0107
Hi there, I'm encountering the following error when executing the Louis method p-value correction method as demonstrated in your tutorial.
In trying to disagnose whats wrong, I notice that the de slot conatins annotations of Type NULL
For the summary statistics, I calculated (using your example from the tutorial) the beta and beta_var vlues from p-value and log2FC from the FindConservedMarkers function from within Seurat.
For the gene specific annotations, I selected the rows from within the humanGeneSets data included with iDEA that corresponded with the genes returned from the DE analysis.
This is the command I used to create the iDEA object:
And this is the command I used to fit the model:
For information, here is my session info:
Could I have some assistance?
The only thing I can think may be problematic is that I have provided a fairly small set of genes, totaling at 68 for this group, as that was the list returned by the Seurat FindConservedMarkers() method. The only other thing I can think of is that when I was calculating the beta and beta_var, my input data as log2FC not logFC. Should I adjust the calculation accoridngly to address this?
Thank you in advance, I am excited to use this tool as I analyze these data! Best, Christian