R/drc_fit.R

In tdanker/ephys2: Read, analyze and plot HEKA Patchmaster files

#' Fit drc models
#'
#' drc_fit expects a data frame with the first three columns represeting response, concentration, and compound (in that order, but names are ignored) and fits drm models for each compound.
#' The output can be passed to \link{drc_plot} or \link{drc_table}.
#'
#' For more info, see \href{../doc/ggDRC.html}{\code{vignette("ggDRC", package = "ggDRC")}}
#'
#' @param df  data frame with data to fit.
#'
#' @param params determines the type of fit
#'
#' @return a tibble with a drc model for each compound
#'
#'
#'
#' @export
drc_fit<-function(df, params=2){
  library(broom)
  df <- dplyr::ungroup(df)
  names(df)<-c("response", "concentration", "compound")

  # define drm function to use with map
  fct=switch(EXPR=params,NULL,drc::LL2.2, drc::LL2.3, drc::LL2.4)
  names=switch(EXPR=params,NULL,
                c("HILL", "IC50"),
                c("HILL", "max","IC50"),
                c("HILL", "min", "max","IC50")
               )

  drm.func <- function(x) {
    drc::drm(response ~ concentration,
        fct = fct(names = names),
        data = x)
  }

  predict.fun <- function(x) {

    modelr::add_predictions(data.frame(x = exp(seq(-20,10, length.out = 100))), x)

  }

  coefs.fun <- function(x) {
    coef(x) %>% tidy -> coefs

    coefs[params,2]<-exp(coefs[params,2])
    coefs
  }

  table.fun <-function(x) {
    x %>% rowwise %>%
      dplyr::mutate(names=names %>% stringr::str_split_fixed(., pattern=":", n=2) %>% .[1]) %>%
      tidyr::spread(names, x)
  }

  df %>% dplyr::group_nest(compound) %>%
    dplyr::mutate(drmod = purrr::map(data, drm.func),
           pred  = purrr::map(drmod, predict.fun),
           coefs = purrr::map(drmod, coefs.fun),
           table = purrr::map(coefs, table.fun))
}