make_ranking_violin_plot: make_ranking_violin_plot

In MonashBioinformaticsPlatform/celaref: Single-cell RNAseq cell cluster labelling by reference

Description

Plot a panel of violin plots showing the distribution of the 'top' genes of each of query group, across the reference dataset.

Usage

make_ranking_violin_plot(de_table.marked = NA, de_table.test = NA,
  de_table.ref = NA, log10trans = FALSE, ...)

Arguments

de_table.marked	The output of get_the_up_genes_for_all_possible_groups for the contrast of interest.
de_table.test	A differential expression table of the query experiment, as generated from contrast_each_group_to_the_rest
de_table.ref	A differential expression table of the reference dataset, as generated from contrast_each_group_to_the_rest
log10trans	Plot on a log scale? Useful for distinishing multiple similar, yet distinct cell type that bunch at top of plot. Default=FALSE.
...	Further options to be passed to get_the_up_genes_for_all_possible_groups, e.g. rankmetric

Details

In the plot output, each panel correponsds to a different group/cluster in the query experiment. The x-axis has the groups in the reference dataset. The y-axis is the rescaled rank of each 'top' gene from the query group, within each reference group.

Only the 'top' genes for each query group are plotted, forming the violin plots - each individual gene is shown as a tickmark. Some groups have few top genes, and so their uncertanty can be seen on this plot.

The thick black lines reprenset the median gene rescaled ranking for each query group / reference group combination. Having this fall above the dotted median threshold marker is a quick indication of potential similarity. A complete lack of similarity would have a median rank around 0.5. Median rankings much less than 0.5 are common though (an 'anti-cell-groupA' signature), because genes overrepresented in one group in an experiment, are likely to be relatively 'underrepresented' in the other groups. Taken to an extreme, if there are only two reference groups, they'll be complete opposites.

Input can be either the precomputed de_table.marked object for the comparison, OR both de_table.test and de_table.ref differential expression results to compare from contrast_each_group_to_the_rest

Value

A ggplot object.

See Also

get_the_up_genes_for_all_possible_groups To make the input data.

Examples

# Make input
# de_table.demo_query <- contrast_each_group_to_the_rest(demo_query_se, "demo_query")
# de_table.demo_ref   <- contrast_each_group_to_the_rest(demo_ref_se,   "demo_ref")
   
# This:                                                  
make_ranking_violin_plot(de_table.test=de_table.demo_query, 
                         de_table.ref=de_table.demo_ref ) 
                        
# Is equivalent to this:
de_table.marked.query_vs_ref <- 
     get_the_up_genes_for_all_possible_groups( de_table.test=de_table.demo_query, 
                                               de_table.ref=de_table.demo_ref)
make_ranking_violin_plot(de_table.marked.query_vs_ref)

MonashBioinformaticsPlatform/celaref documentation built on June 5, 2019, 11:35 a.m.