R/topslam.R

In dynverse/topslam: Wrapper for topslam

#' Run topslam
#'
#' @param expression the normalised expression data
#' @param start_cell_id the names of starting cells
#' @param n_components the number of components
#' @param n_neighbors the number of neighbors
#' @param linear_dims ?
#' @param max_iters the number of iterations to optimize over
#' @param dimreds the dimensionality reductions to use
#' @param num_cores the number of cores
#'
#' @importFrom methods formalArgs
#' @importFrom utils write.table read.csv
#' @importFrom jsonlite toJSON
#' @export
topslam <- function(
  expression,
  start_cell_id,
  n_components = 2,
  n_neighbors = 10,
  linear_dims = 0,
  max_iters = 200,
  dimreds = c("t-SNE", "PCA", "Spectral", "Isomap", "ICA"),
  num_cores = 1
) {
  # python counts from 0, R from 1
  start_cell_id <- which(rownames(expression) %in% start_cell_id) - 1

  # create a temporary folder
  temp_folder <- tempfile()
  dir.create(temp_folder, recursive = TRUE)

  tryCatch({
    # write expression data
    utils::write.table(as.data.frame(expression), paste0(temp_folder, "/expression.tsv"), sep="\t")

    # write params to json
    copy_args <- setdiff(methods::formalArgs(topslam), c("expression"))
    params <- as.list(environment())[copy_args]
    write(jsonlite::toJSON(params, auto_unbox = TRUE), paste0(temp_folder, "/params.json"))

    if (!is.null(num_cores)) {
      num_cores_str <- glue::glue(
        "export MKL_NUM_THREADS={num_cores};",
        "export NUMEXPR_NUM_THREADS={num_cores};",
        "export OMP_NUM_THREADS={num_cores}"
      )
    } else {
      num_cores_str <- "echo hi"
    }

    commands <- glue::glue(
      "cd {find.package('topslam')}",
      "source venv/bin/activate",
      "{num_cores_str}",
      "cd ..",
      "python {find.package('topslam')}/wrapper.py {temp_folder}",
      .sep = ";"
    )
    output <- processx::run("/bin/bash", c("-c", commands), echo=TRUE)

    # read output
    wad_grid <- utils::read.csv(paste0(temp_folder, "/wad_grid.csv"))
    wad_energy <- utils::read.csv(paste0(temp_folder, "/wad_energy.csv"))
    space <- utils::read.csv(paste0(temp_folder, "/space.csv"))
    pseudotime <- utils::read.csv(paste0(temp_folder, "/pseudotime.csv"))
  }, finally = {
    # remove temporary output
    unlink(temp_folder, recursive = TRUE)
  })

  # return output
  list(
    wad_grid = wad_grid,
    wad_energy = wad_energy,
    space = space,
    pseudotime = pseudotime
  )
}