In Bioconductor/SOUNDBoard: Statistically-informed Multi-Omics Reports for Tumor and Other Boards

knitr::opts_knit$set(root.dir =
    file.path(SOUNDBoard:::.options$get("resources"), "html"))

This vignette provides a quick tour of SOUNDBoard. There are three steps to creating a SOUND Board Report: assembly, presentation, and deployment

Basic use

Assembly

Load the SOUNDBoard package

suppressPackageStartupMessages({
    library(SOUNDBoard)
    library(shiny)
})

Create a SOUNDManager(). The first argument is the path to the local version of your report and its assets. We use a temporary location for the vignette. The second (optional) argument is a file defining SQL tables board, cases, and assays via schemas. The schemas are used to initialize the data base used to store board information.

x <- SOUNDManager(tempfile())
x

Now create records corresponding to boards and cases you are tracking.

tbl(x, "board") <- list(
    board_uid = "MitoNET",
    description = "Clinical management of rare genetic disorders"
)

tbl(x, "cases") <- list(
    board_uid = "MitoNET",
    case_uid = "PATIENT_1",
    age = 36,
    sex = "Male"
)

Add 'assays' for cases under investigation. Each assay should be a single R object, file, etc. Use manage() to serialize the object. We provide two examples, to illustrate assays that might be summarized as data frames or images.

tbl(x, "assay") <- list(
    case_uid = "PATIENT_1",
    assay = "RNA candidates",
    description = "RNA candidate genes derived from analysis ...",
    resource = manage(x, data.frame())
)

tbl(x, "assay") <- list(
    case_uid = "PATIENT_1",
    assay = "data.frame",
    description = "Demonstrating data frame management",
    resource = manage(x, mtcars)
)

tbl(x, "assay") <- list(
    case_uid = "PATIENT_1",
    assay = "ggplot",
    description = "Demonstrating ggplot management",
    resource = manage(x, ggplot2::qplot(hp, mpg, data=mtcars))
)

Assays can be complicated, e.g., an entire shiny application:

## create a shiny application

myapp <- ShinyAppWidget(
    shinyApp(
        ui = bootstrapPage(
            numericInput('n', 'Number of obs', 100),
            plotOutput('plot')
        ),
        server = function(input, output) {
            output$plot <- renderPlot({
                hist(runif(input$n))
            })
        }
    )
)

Add the shiny app as a managed ShinyWidget() assay

tbl(x, "assay") <- list(
    case_uid = "PATIENT_1",
    assay = "shiny",
    description = "Demonstrating shiny application",
    resource = manage(x, myapp)
)

Review the information under management

x
tbl(x, "board")
tbl(x, "cases")
tbl(x, "assay")

Report preparation

Prepare a report by extracting relevant rows from the board, cases, or assay tables, and present them in a markdown document with sbreport(). The code chunks would typically include echo=FALSE. Here we present the code chunks with eval=FALSE, and then again with echo=FALSE to see the results in this document.

Approximate content

'''{r}
tbl(x, "board") %>% filter(board_uid == "MitoNET") %>% sbreport()

tbl(x, "cases") %>% filter(case_uid == "PATIENT_1") %>%
    select(-board_uid) %>% sbreport()
'''

'''{r}
tbl(x, "assay") %>%
    filter(case_uid == "PATIENT_1", assay == "data.frame") %>%
    sbreport()
'''

'''{r}
tbl(x, "assay") %>%
    filter(case_uid == "PATIENT_1", assay == "ggplot") %>%
    sbreport()
'''

'''{r}
tbl(x, "assay") %>%
    filter(case_uid == "PATIENT_1", assay == "shiny") %>%
    sbreport()
'''

Approximate output

tbl(x, "board") %>% filter(board_uid == "MitoNET") %>% sbreport()

tbl(x, "cases") %>% filter(case_uid == "PATIENT_1") %>%
    select(-board_uid) %>% sbreport()

Demonstrating plot (ggplot2 / plotly) management

tbl(x, "assay") %>%
    filter(case_uid == "PATIENT_1", assay == "ggplot") %>%
    sbreport()

Demonstrating data frame (data.frame() / DT) management

tbl(x, "assay") %>%
    filter(case_uid == "PATIENT_1", assay == "data.frame") %>%
    sbreport()

Embedded apps: more advanced calculations can call back into R.

tbl(x, "assay") %>%
    filter(case_uid == "PATIENT_1", assay == "shiny") %>%
    sbreport()

Deployment

manager <- SOUNDManager(
    x,
    host = "localhost", port = "3838", path = "SOUNDBoard",
    username = "soundboard"
)

deploy(manager)

Extending reports

Defining a SOUNDWidget

The SOUNDWidget() function allows users to create alternative ways to store, load, and report different data types. The argument requires a single argument, the name of the widget, and three optional arguments. The optional arguments are functions that describe how the content of the widget is to be saved to disk (e.g., saveRDS()), loaded from disk (e.g., readRDS()) and presented in the report (e.g., DT::datatable()). Here we create a simple widget that manages a data.frame(), reporting it as an interactive DT datatable() (this behavior is provided by default; the widget is developed for illustration only).

DTWidget <- SOUNDWidget(
    "DTWidget",
    save = function(x, file) saveRDS(x, file),
    load = function(x, file) readRDS(file),
    report = function(x) DT::datatable(sbresource(x))
)

The SOUNDWidget() function returns a constructor, to be used with the object to be manipulated, e.g., using the USArrests built-in data set.

arrests <- DTWidget(USArrests)
arrests
head(sbresource(arrests))

Testing

The functionality can be tested directly

fl <- tempfile()
sbsave(arrests, fl)
reloaded <- sbload(DTWidget(), fl)

Running the sbreport() command generates an html page containing the USArrests data set as an interative html-based table.

sbreport(reloaded)

Use

The new widget can be incorporated into a report as above, e.g., adding to the data base

tbl(x, "assay") <- list(
    case_uid = "PATIENT_1",
    assay = "arrests",
    description = "Demonstrating custom widget management",
    resource = manage(x, arrests)
)

or including in the report

tbl(x, "assay") %>% filter(assay == "arrests") %>% sbreport()

Bioconductor/SOUNDBoard documentation built on May 6, 2019, 7:52 a.m.