17-plotting_repeat_evals.Rmd

---
title: "Plotting for the repeated SLE WB and subsampled recount2 model evaluations"
output:   
  html_notebook: 
    toc: true
    toc_float: true
---

**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

```{r}
`%>%` <- dplyr::`%>%`
```

```{r}
# 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

```{r}
sle.sparsity <- readr::read_tsv(file.path("results", "16",
                                          "sle-wb_repeated_sparsity.tsv")) %>%
                  dplyr::mutate(training_set = "SLE")
sle.num.lvs <- readr::read_tsv(file.path("results", "16",
                                          "sle-wb_repeated_num_lvs.tsv")) %>%
                  dplyr::mutate(training_set = "SLE")
sle.pathway <- readr::read_tsv(file.path("results", "16",
                                          "sle-wb_repeated_pathway.tsv")) %>%
                  dplyr::mutate(training_set = "SLE")
```

### Subsampled recount2

```{r}
recount.sparsity <- readr::read_tsv(file.path("results", "15",
                                          "subsampled_sparsity.tsv")) %>%
                  dplyr::mutate(training_set = "recount2 subsampled")
recount.num.lvs <- readr::read_tsv(file.path("results", "15",
                                          "subsampled_num_lvs.tsv")) %>%
                  dplyr::mutate(training_set = "recount2 subsampled")
recount.pathway <- readr::read_tsv(file.path("results", "15",
                                          "subsampled_pathway.tsv")) %>%
                  dplyr::mutate(training_set = "recount2 subsampled")
```

### Bind two experiments

```{r}
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

```{r}
# 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

```{r}
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")
```

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

#### Pathway coverage

```{r}
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")
```

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

#### `U` sparsity

```{r}
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")
```


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