---
title: "Reading and writing data"
author: "Timothy Keyes"
date: "`r Sys.Date()`"
output: rmarkdown::html_vignette
description: > 
  Read this vignette to learn how to read data from FCS and CSV files
  on disk and how to write data to FCS and CSV files on disk. 
vignette: >
  %\VignetteIndexEntry{02. Reading and writing data}
  %\VignetteEngine{knitr::knitr}
  %\VignetteEncoding{UTF-8}
---

```{r, include = FALSE}
knitr::opts_chunk$set(
    collapse = TRUE,
    comment = "#>"
)

options(
  tibble.print_min = 4L, 
  tibble.print_max = 4L, 
  rmarkdown.html_vignette.check_title = FALSE
)
```

```{r setup, message = FALSE}
library(tidytof)
library(dplyr)
```

This vignette teaches you how to read CyTOF data into an R session from two common file formats in which CyTOF data is typically stored: Flow Cytometry Standard (FCS) and Comma-Separated Value (CSV) files.

## Accessing the data for this vignette

`{tidytof}` comes bundled with several example mass cytometry datasets. To access the raw FCS and CSV files containing these data, use the `tidytof_example_data` function. When called with no arguments, `tidytof_example_data` will return a character vector naming the datasets contained in `{tidytof}`:

```{r}
tidytof_example_data()
```

The details of the datasets contained in each of these directories isn't particularly important, but some basic information is as follows:

-   **aml** - one FCS file containing myeloid cells from a healthy bone marrow and one FCS file containing myeloid cells from an AML patient bone marrow
-   **ddpr** - two FCS files containing B-cell lineage cells from [this paper](https://pubmed.ncbi.nlm.nih.gov/29505032/)
-   **mix** - two FCS files with different CyTOF antigen panels (one FCS file from the "aml" directory and one from the "phenograph" directory)
-   **mix2** - three files with different CyTOF antigen panels and different file extensions (one FCS file from the "aml" directory and two CSV files from the "phenograph_csv directory)
-   **phenograph** - three FCS files containing AML cells from [this paper](https://pubmed.ncbi.nlm.nih.gov/26095251/)
-   **phenograph_csv** - the same cells as in the "phenograph" directory, but stored in CSV files
-   **scaffold -** three FCS files from [this paper](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4537647/)
-   **statistical_scaffold** - three FCS files from [this paper](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5312823/)
-   **surgery** - three FCS files from [this paper](https://pubmed.ncbi.nlm.nih.gov/25253674/)

To obtain the file path for the directory containing each dataset, call `tidytof_example_data` with one of these dataset names as its argument. For example, to obtain the directory for the phenograph data, we would use the following command:

```{r}
tidytof_example_data("phenograph")
```

## Reading Data with `tof_read_data`

Using one of these directories (or any other directory containing CyTOF data on your local machine), we can use `tof_read_data` to read CyTOF data from raw files. Acceptable formats include FCS files and CSV files. Importantly, `tof_read_data` is smart enough to read single FCS/CSV files or multiple FCS/CSV files depending on whether its first argument (`path`) leads to a single file or to a directory of files.

Here, we can use `tof_read_data` to read in all of the FCS files in the "phenograph" example dataset bundled into `{tidytof}` and store it in the `phenograph` variable.

```{r}
phenograph <-
    tidytof_example_data("phenograph") %>%
    tof_read_data()

phenograph %>%
    class()
```

Regardless of the input data file type, `{tidytof}` reads data into an extended `tibble` class called a `tof_tbl` (pronounced "tof tibble"). 

tof tibbles are an S3 class identical to `tbl_df`, but with one additional attribute ("panel"). `{tidytof}` stores this additional attribute in `tof_tbl`s because, in addition to analyzing CyTOF data from individual experiments, CyTOF users often want to compare panels between experiments to find common markers or to compare which metals are associated with particular markers across panels. To retrieve this panel information from a `tof_tbl`, use `tof_get_panel`: 

```{r}
phenograph %>%
    tof_get_panel()
```

A few additional notes about `tof_tbl`s:

-   `tof_tbl`s contains one cell per row and one CyTOF channel per column (to provide the data in its "tidy" format).
-   `tof_read_data` adds an additional column to the output `tof_tbl` encoding the name of the file from which each cell was read (the "file_name" column).
- Because `tof_tbl`s inherit from the `tbl_df` class, all methods available to tibbles are also available to `tof_tbl`s. 

## Using tibble methods with `{tidytof}` tibbles

As an extension of the `tbl_df` class, `tof_tbl`s get access to all `{dplyr}` and `{tidyr}` for free. These can be useful for performing a variety of common operations.

For example, the `phenograph` object above has two columns - `PhenoGraph` and `Condition` - that encode categorical variables as numeric codes. We might be interested in converting the types of these columns into strings to make sure that we don't accidentally perform any quantitative operations on them later. Thus, `{dplyr}`'s useful `mutate` method can be applied to `phenograph` to convert those two columns into character vectors.

```{r}
phenograph <-
    phenograph %>%
    # mutate the input tof_tbl
    mutate(
        PhenoGraph = as.character(PhenoGraph),
        Condition = as.character(Condition)
    )

phenograph %>%
    # use dplyr's select method to show
    # that the columns have been changed
    select(where(is.character))
```

And note that the `tof_tbl` class is preserved even after these transformations.

```{r}
phenograph %>%
    class()
```


Importantly, `tof_read_data` uses an opinionated heuristic to mine different keyword slots of input FCS file(s) and guess which metals and antigens were used during data acquisition. Thus, when CSV files are read using `tof_read_data`, it is recommended to use the `panel_info` argument to provide the panel manually (as CSV files, unlike FCS files, do not provide built-in metadata about the columns they contain).

```{r}
# when csv files are read, the tof_tibble's "panel"
# attribute will be empty by default
tidytof_example_data("phenograph_csv") %>%
    tof_read_data() %>%
    tof_get_panel()

# to add a panel manually, provide it as a tibble
# to tof_read_data
phenograph_panel <-
    phenograph %>%
    tof_get_panel()

tidytof_example_data("phenograph_csv") %>%
    tof_read_data(panel_info = phenograph_panel) %>%
    tof_get_panel()
```


## Writing data from a `tof_tbl` to disk

Users may wish to store CyTOF data as FCS or CSV files after transformation, concatenation, filtering, or other data processing. To write single-cell data from a `tof_tbl` into FCS or CSV files, use `tof_write_data`. To illustrate how to use this verb, we use the `{tidytof}`'s built-in `phenograph_data` dataset.

```{r}
data(phenograph_data)

print(phenograph_data)
```

```{r, eval = FALSE}
# when copying and pasting this code, feel free to change this path
# to wherever you'd like to save your output files
my_path <- file.path("~", "Desktop", "tidytof_vignette_files")

phenograph_data %>%
    tof_write_data(
        group_cols = phenograph_cluster,
        out_path = my_path,
        format = "fcs"
    )
```

`tof_write_data`'s trickiest argument is `group_cols`, the argument used to specify which columns in `tof_tibble` should be used to group cells (the rows of `tof_tibble`) into separate FCS or CSV files. Simply put, this argument allows `tof_write_data` to create a single FCS or CSV file for each unique combination of values in the `group_cols` columns specified by the user. In the example above, cells are grouped into 3 output FCS files - one for each of the 3 clusters encoded by the `phenograph_cluster` column in `phenograph_data`. These files should have the following names (derived from the values in the `phenograph_cluster` column):

-   cluster1.fcs
-   cluster2.fcs
-   cluster3.fcs

Note that these file names match the distinct values in our `group_cols` column (`phenograph_cluster`): 

```{r}
phenograph_data %>%
    distinct(phenograph_cluster)
```

However, suppose we wanted to write multiple files for each cluster by breaking cells into two groups: those that express high levels of `pstat5` and those that express low levels of `pstat5`. We can use `dplyr::mutate` to create a new column in `phenograph_data` that breaks cells into high- and low-`pstat5` expression groups, then add this column to our `group_cols` specification:

```{r, eval = FALSE}
phenograph_data %>%
    # create a variable representing if a cell is above or below
    # the median expression level of pstat5
    mutate(
        expression_group = if_else(pstat5 > median(pstat5), "high", "low")
    ) %>%
    tof_write_data(
        group_cols = c(phenograph_cluster, expression_group),
        out_path = my_path,
        format = "fcs"
    )
```

This will write 6 files with the following names (derived from the values in `phenograph_cluster` and `expression_group`).

-   cluster1_low.fcs
-   cluster1_high.fcs
-   cluster2_low.fcs
-   cluster2_high.fcs
-   cluster3_low.fcs
-   cluster3_high.fcs

As above, note that these file names match the distinct values in our `group_cols` columns (`phenograph_cluster` and `expression_group`):

```{r}
phenograph_data %>%
    mutate(
        expression_group = if_else(pstat5 > median(pstat5), "high", "low")
    ) %>%
    distinct(phenograph_cluster, expression_group)
```

A useful feature of `tof_write_data` is that it will automatically concatenate cells into single FCS or CSV files based on the specified `group_cols` *regardless of how many unique files those cells came from*. This allows for easy concatenation of FCS or CSV files containing data from a single sample acquired over multiple CyTOF runs, for example.

# Session info

```{r}
sessionInfo()
```