R/pipeline.R

In wlandau/targets-shiny: Shiny App with a Targets Pipeline

# Write the _targets.R script to the project directory.
write_pipeline <- function(
  name,
  biomarkers =  c("albumin", "log_bilirubin"),
  iterations = 1000
) {
  tar_helper(project_path(name, "_targets.R"), {
    suppressPackageStartupMessages({
      library(dplyr)
      library(targets)
      library(tarchetypes)
      library(tidyr)
    })
    options(cli.num_colors = 1)
    tar_option_set(
      packages = c("ggplot2", "rstanarm", "tibble"),
      memory = "transient",
      garbage_collection = TRUE
    )
    get_data_biomarker <- function() {
      pbcLong %>%
        rename(log_bilirubin = logBili) %>%
        mutate(log_platelet = log(platelet))
    }
    fit_model <- function(data_biomarker, biomarker, iterations) {
      model <- stan_jm(
        formulaLong = as.formula(paste(biomarker, "~ year + (1 | id)")),
        formulaEvent = Surv(futimeYears, death) ~ sex + trt,
        time_var = "year",
        dataLong = data_biomarker,
        dataEvent = pbcSurv,
        iter = iterations,
        refresh = min(iterations / 10, 50)
      )
      tibble(
        alpha = as.data.frame(model)[["Assoc|Long1|etavalue"]],
        biomarker = biomarker
      )
    }
    plot_samples <- function(samples) {
      ggplot(samples) +
        geom_density(aes(x = alpha, fill = biomarker), alpha = 0.5) +
        theme_gray(20)
    }
    models <- tar_map(
      values = list(biomarker = !!sort(biomarkers)),
      tar_fst_tbl(model, fit_model(data_biomarker, biomarker, !!iterations))
    )
    list(
      tar_target(data_biomarker, get_data_biomarker()),
      models,
      tar_combine(samples, models),
      tar_qs(plot, plot_samples(samples))
    )
  })
}