pretty_MA_plot doesn't label genes if convert_ids=FALSE

[luisa-cusick]

1. Need to cast param results to data.frame (if convert_ids=TRUE this happens in the id_convert function). For a DESeq results object, this just strips all classes preceding "data.frame" so that ggplot recognizes it as a data frame
2. If name_col is the default ('row.names'), geom_lable_repel layers fail in label assignment because name_col needs to be a column name:
   geom_label_repel(data = subset(results, name_col %in% genes & log2FoldChange > 0), aes(label = .data[[name_col]], size = factor(name_col %in% highlight_genes)), min.segment.length = 0.1, max.overlaps = max_overlaps, fill = alpha(c("white"), label_alpha), ylim = c(1, NA))

[RoganGrant]

Thank you for raising this, Luisa! Many of these functions could honestly use some housekeeping. Just so I fully understand, what is the use case for setting convert_ids=FALSE? Pseudo-bulk?

[luisa-cusick]

Use case for convert_ids=FALSE is when aligning to a genome that does not use ensemble gene IDs.

Fixed the issue on this branch, submitting PR

[RoganGrant]

Need to roll this back because it will break a lot of workflows. The use of name_col here is problematic, as it's a character vector of length 1. So for instance on line 126: size = factor(name_col %in% highlight_genes)) will inevitably always be false. All of the highlights and labels are therefore lost. I am rolling back to the previous commit and we can either make an alt version or try a different solution.