Does Numbat filter cells out during giving posterior probabilities on single-cell level?

[aelhossiny]

Hi, I have noticed that some of the cells are not reported in the joint_post slot even though they present in the original bam file. I might have missed it somewhere in the paper or the documentation, but do you filter out cells during computing posterior probabilities on single-cell level?

Thanks

[teng-gao]

Hi,

The only filtering for cells are here:

numbat/R/main.R

Lines 106 to 117 in a367fa5

	zero_cov = names(which(colSums(count_mat) == 0))
	if (length(zero_cov) > 0) {
	log_message(glue('Filtering out {length(zero_cov)} cells with 0 coverage'))
	count_mat = count_mat[,!colnames(count_mat) %in% zero_cov]
	df_allele = df_allele %>% filter(!cell %in% zero_cov)
	}
	# only keep cells that have a transcriptome
	df_allele = df_allele %>% filter(cell %in% colnames(count_mat))
	if (nrow(df_allele) == 0){
	stop('No matching cell names between count_mat and df_allele')
	}

So, all cells in count_mat should be present in joint_post unless they have zero coverage.