`mark_dominant()` deletes `normalized_score` in `experiment@counts$TSSs$raw[sample]`

[perinom]

Hi,

I just realised that setting use_nomalized = TRUE in plot_seqeunce_logo() results in the following error after TSSs aggregation into TSRs:

Error in eval(jsub, SDenv, parent.frame()) : object 'normalized_score' not found

I traced it back to line 34 of extract_counts() where the function tries to extract the normalized_score column that is indeed not present. The score column is present but contains floating point numbers instead of integers. I guess it went through some kind of processing, but it does not coincide with the normalized_score calculated on TSSs before

As the error does not occur when only TSSs are present in the experiment object I checked the functions in between and this seems to happen during the call to mark_dominant()

------- EDIT -------
it only happens when setting use_normalized = T in mark_dominant()

Full traceback:

13. eval(jsub, SDenv, parent.frame())
12. eval(jsub, SDenv, parent.frame())
11. `[.data.table`(x, , `:=`(score, normalized_score))
10. x[, `:=`(score, normalized_score)]
9.  .f(.x[[i]], ...)
8. map(.x, .f, ...)
7. walk(return_samples, function(x) {
    x[, `:=`(score, normalized_score)]
    x[, `:=`(normalized_score, NULL)]
    return(x) ...
6. extract_counts(., "tss", samples, use_normalized)
5. map(signal_data, function(x) {
    x <- x[score >= threshold]
    return(x)
    })
4. preliminary_filter(., dominant, threshold)
3. experiment %>% extract_counts("tss", samples, use_normalized) %>%
    preliminary_filter(dominant, threshold)
2. .tss_sequences(experiment, samples, genome_assembly, threshold,
    use_normalized, distance, dominant, data_conditions)
1. plot_sequence_logo(exp, samples = "all", genome_assembly = assembly,
    threshold = threshold, use_normalized = TRUE, distance = 6,
    dominant = FALSE, data_conditions = NULL, ncol = 3, font_size = 10,
    base_colors = c(A = "#109649", C = "#255C99", G = "#F7B32C", ...