Plotting for the repeated SLE WB and subsampled recount2 model evaluations

J. Taroni 2018

In 15-evaluate_subsampling and 16-repeat_sle_wb_PLIER.R, we calculated some measures of PLIER model performance (e.g., sparsity of U). We’ll plot the results for those two sets of training data in this notebook.

Functions and directory setup

`%>%` <- dplyr::`%>%`

# plot and result directory setup for this notebook
plot.dir <- file.path("plots", "17")
dir.create(plot.dir, recursive = TRUE, showWarnings = FALSE)
results.dir <- file.path("results", "17")
dir.create(results.dir, recursive = TRUE, showWarnings = FALSE)

Read in data

Read in the three data.frame from each set of experiments: sparsity, pathway coverage, and number of LVs.

SLE WB repeats

sle.sparsity <- readr::read_tsv(file.path("results", "16",
                                          "sle-wb_repeated_sparsity.tsv")) %>%
                  dplyr::mutate(training_set = "SLE")

Parsed with column specification:
cols(
  model = col_character(),
  sparsity_type = col_character(),
  value = col_double()
)

sle.num.lvs <- readr::read_tsv(file.path("results", "16",
                                          "sle-wb_repeated_num_lvs.tsv")) %>%
                  dplyr::mutate(training_set = "SLE")

Parsed with column specification:
cols(
  model = col_character(),
  num_lvs = col_integer()
)

sle.pathway <- readr::read_tsv(file.path("results", "16",
                                          "sle-wb_repeated_pathway.tsv")) %>%
                  dplyr::mutate(training_set = "SLE")

Parsed with column specification:
cols(
  model = col_character(),
  pathway_coverage_type = col_character(),
  value = col_double()
)

Subsampled recount2

recount.sparsity <- readr::read_tsv(file.path("results", "15",
                                          "subsampled_sparsity.tsv")) %>%
                  dplyr::mutate(training_set = "recount2 subsampled")

Parsed with column specification:
cols(
  model = col_character(),
  sparsity_type = col_character(),
  value = col_double()
)

recount.num.lvs <- readr::read_tsv(file.path("results", "15",
                                          "subsampled_num_lvs.tsv")) %>%
                  dplyr::mutate(training_set = "recount2 subsampled")

Parsed with column specification:
cols(
  model = col_character(),
  num_lvs = col_integer()
)

recount.pathway <- readr::read_tsv(file.path("results", "15",
                                          "subsampled_pathway.tsv")) %>%
                  dplyr::mutate(training_set = "recount2 subsampled")

Parsed with column specification:
cols(
  model = col_character(),
  pathway_coverage_type = col_character(),
  value = col_double()
)

Bind two experiments

num.lvs.df <- dplyr::bind_rows(sle.num.lvs, recount.num.lvs)
pathway.df <- dplyr::bind_rows(sle.pathway, recount.pathway)
sparsity.df <- dplyr::bind_rows(sle.sparsity, recount.sparsity)
rm(sle.num.lvs, sle.pathway, sle.sparsity, recount.num.lvs, recount.pathway,
     recount.sparsity)

Write number of LVs data.frame to file

# number of latent variables
num.lvs.file <- file.path(results.dir, "number_of_LVs.tsv")
readr::write_tsv(num.lvs.df, num.lvs.file)

Plotting

Number of latent variables

num.lvs.df %>%
  ggplot2::ggplot(ggplot2::aes(x = training_set, y = num_lvs,
                               group = training_set)) +
  ggplot2::geom_boxplot() +
  ggplot2::geom_point(position = ggplot2::position_jitter(0.2),
                      alpha = 0.5) +
  ggplot2::theme_bw() +
  ggplot2::labs(x = "training set", 
                y = "number of latent variables",
                title = "PLIER model n = 1640")

plot.file <- file.path(plot.dir, "number_of_lvs.pdf")
ggplot2::ggsave(plot.file, plot = ggplot2::last_plot())

Saving 7 x 7 in image

Pathway coverage

pathway.df <- pathway.df %>%
  dplyr::filter(pathway_coverage_type != "sig.pathway.by.lv") %>%
  dplyr::mutate(pathway_coverage_type = 
                  dplyr::case_when(
                    (pathway_coverage_type == "lv") ~ 
                      "LV associated with pathways",
                    (pathway_coverage_type =="pathway") ~ "pathway coverage"
                  ))
# pathway coverage
pathway.file <- file.path(results.dir, "pathway_coverage.tsv")
readr::write_tsv(pathway.df, pathway.file)
pathway.df %>%
  ggplot2::ggplot(ggplot2::aes(x = training_set, y = value)) +
  ggplot2::geom_boxplot() +
  ggplot2::geom_point(position = ggplot2::position_jitter(0.2),
                      alpha = 0.5) +
  ggplot2::facet_grid(~ pathway_coverage_type) +
  ggplot2::theme_bw() +
  ggplot2::labs(x = "training set", 
                y = "proportion",
                title = "PLIER model n = 1640")

plot.file <- file.path(plot.dir, "pathway_coverage.pdf")
ggplot2::ggsave(plot.file, plot = ggplot2::last_plot())

Saving 7 x 7 in image

U sparsity

sparsity.df %>%
  dplyr::mutate(sparsity_type = 
                dplyr::case_when(
                  (sparsity_type == "all.sparsity") ~ "All",
                  (sparsity_type == "sig.sparsity") ~ "Significant association only"
                )) %>%
  ggplot2::ggplot(ggplot2::aes(x = training_set, y = value)) +
  ggplot2::geom_violin() +
  ggplot2::stat_summary(fun.y = median, geom = "point", shape = 18,
                        size = 4) +
  ggplot2::facet_grid(~ sparsity_type) +
  ggplot2::theme_bw() +
  ggplot2::labs(x = "training set", 
                y = "proportion of positive entries in U",
                title = "PLIER model n = 1640")

plot.file <- file.path(plot.dir, "u_sparsity.pdf")
ggplot2::ggsave(plot.file, plot = ggplot2::last_plot())

Saving 7 x 7 in image