jj_get_pval_df: get p values for group comparisons with wilcox test

In mathosi/jj: Just Joise

get p values for group comparisons with wilcox test

Description

get a data.frame of pvalues which can be passed to ggpubr to include in a barplot/boxplot... Each donor can have multiple samples. Each donor should uniquely belong to one group

Usage

jj_get_pval_df(
  df,
  sample_column,
  donor_column,
  group_column,
  score_column,
  comparisons_paired,
  comparisons_unpaired,
  p_adj_method = "BH",
  p_adj_round = 3,
  return_per_sample_df = FALSE
)

Arguments

sample_column	column containing annotation for the sample from which the observations are derived. number of columns
donor_column	column specifying the donor from which the observations are derived. If all samples are unpaired, it is the same as sample_column but must be passed as a second column to dplyr number of rows
group_column	column specifying the groups that should be used for the statistical comparison
score_column	column containing the values that are used for the comparison between groups (wilcox.test)
comparisons_paired	list of vectors of length 2 specifying the desired paired comparisons between levels in the group_column
comparisons_unpaired	list of vectors of length 2 specifying the desired unpaired comparisons between levels in the group_column

Value

Returns a tibble of pvalues

Examples

#make some example data: 10 samples, 4 healthy donors and 3 patients. From each patients, values are available from
#two timepoints: diagnosis and relapse. We want to quantify the difference in a `Disease_score` between Healthy, Diagnosis and 
#relapse state on the sample level (not on single-cell level). Both paired (Diagnosis - Relapse) and unpaired
#(Healthy-Diagnosis, Healthy-Relapse) comparisons need to be performed.
data_df = data.frame(Sample = rep(1:10, each = 10),
                     Donor = c(rep(c('Healthy1','Healthy2','Healthy3','Healthy4'), each=10), rep(c('Disease1','Disease2','Disease3'), each=20)),
                     Group = c(rep('Healthy', 40), rep(rep(c('Diagnosis','Relapse'), each=10), 3)),
                     Disease_score = c(rnorm(40, mean = 3, sd = 0.5),  rnorm(20, mean = 4, sd = 0.1),
                                       rnorm(10, mean = 6, sd = 0.1), rnorm(10, mean = 4.5, sd = 0.1),
                                       rnorm(10, mean = 4.5, sd = 0.1), rnorm(10, mean = 3, sd = 0.1)))
#plot the data per group as violins
gg = jj_plot_numeric_by_group(data_df, feature_column = 'Disease_score', group_column = 'Group')
gg
#get mean score per sample
sample_df = jj_get_pval_df(data_df,
                      sample_column = 'Sample',
                      donor_column =  'Donor',
                      group_column = 'Group',
                      score_column =  'Disease_score',
                      return_per_sample_df = T)
#i.e. comparisons which are calculated in the function 'get_pval_df'
wilcox.test(x=sample_df$score[sample_df$group == 'Healthy'], 
            y=sample_df$score[sample_df$group == 'Diagnosis'],
            paired = F)
wilcox.test(x=sample_df$score[sample_df$group == 'Relapse'], 
            y=sample_df$score[sample_df$group == 'Diagnosis'],
            paired = T)
#get the full pvalue tibble
pval_df = jj_get_pval_df(data_df,
                      sample_column = 'Sample',
                      donor_column =  'Donor',
                      group_column = 'Group',
                      score_column =  'Disease_score',
                      comparisons_paired = list(c('Diagnosis','Relapse')),
                      comparisons_unpaired = list(c('Healthy', 'Diagnosis'), c('Healthy', 'Relapse')),
                      )
pval_df
#add suitable y positions for p value visualization in the violin plot
pval_df$ypos = c(7, 6.5, 6.75)
#plot everything together using ggpubr package to show pvalues
gg + geom_point(data = sample_df, mapping = aes(x = group, y = score, colour=donor), size=3) + 
  scale_colour_manual(values=jj_get_jj_colours(sample_df$donor)) +
  ggpubr::stat_pvalue_manual(data = pval_df, label='p_val_adj',
                                tip.length = 0.01,
                                y.position = 'ypos') 

mathosi/jj documentation built on Feb. 25, 2024, 2:29 p.m.