bootstrapNullAlternativeModelFast: Bootstrap null distribution of F statistics for FDR...

View source: R/null_bootstrap_funcs.R

bootstrapNullAlternativeModelFastR Documentation

Bootstrap null distribution of F statistics for FDR estimation based on resampling alternative model residuals with only one round of model fitting on resampled data and subsequent resampling of thereby obtained residuals

Description

Bootstrap null distribution of F statistics for FDR estimation based on resampling alternative model residuals with only one round of model fitting on resampled data and subsequent resampling of thereby obtained residuals

Usage

bootstrapNullAlternativeModelFast(
  df,
  params_df,
  maxit = 500,
  independentFiltering = FALSE,
  fcThres = 1.5,
  minObs = 20,
  optim_fun_h0 = TPP2D:::.min_RSS_h0,
  optim_fun_h1 = TPP2D:::.min_RSS_h1_slope_pEC50,
  optim_fun_h1_2 = NULL,
  gr_fun_h0 = NULL,
  gr_fun_h1 = NULL,
  gr_fun_h1_2 = NULL,
  BPPARAM = BiocParallel::SerialParam(progressbar = TRUE),
  B = 20,
  byMsExp = TRUE,
  verbose = FALSE
)

Arguments

df

tidy data frame retrieved after import of a 2D-TPP dataset, potential filtering and addition of a column "nObs" containing the number of observations per protein

params_df

data frame listing all null and alternative model parameters as obtained by 'getModelParamsDf'

maxit

maximal number of iterations the optimization should be given, default is set to 500

independentFiltering

boolean flag indicating whether independent filtering should be performed based on minimal fold changes per protein profile

fcThres

numeric value of minimal fold change (or inverse fold change) a protein has to show to be kept upon independent filtering

minObs

numeric value of minimal number of observations that should be required per protein

optim_fun_h0

optimization function that should be used for fitting the H0 model

optim_fun_h1

optimization function that should be used for fitting the H1 model

optim_fun_h1_2

optional additional optimization function that will be run with paramters retrieved from optim_fun_h1 and should be used for fitting the H1 model with the trimmed sum model, default is NULL

gr_fun_h0

optional gradient function for optim_fun_h0, default is NULL

gr_fun_h1

optional gradient function for optim_fun_h1, default is NULL

gr_fun_h1_2

optional gradient function for optim_fun_h1_2, default is NULL

BPPARAM

BiocParallel parameter for optional parallelization of null distribution generation through bootstrapping, default: BiocParallel::SerialParam()

B

numeric value of rounds of bootstrap, default: 20

byMsExp

boolean flag indicating whether resampling of residuals should be performed separately for data generated by different MS experiments, default TRUE, recommended

verbose

logical indicating whether to print each protein while its profile is boostrapped

Value

data frame containing F statistics of proteins with permuted 2D thermal profiles that are informative on the Null distribution of F statistics

Examples

data("simulated_cell_extract_df")
temp_df <- simulated_cell_extract_df %>% 
  filter(clustername %in% paste0("protein", 1:3)) %>% 
  group_by(representative) %>% 
  mutate(nObs = n()) %>% 
  ungroup 
temp_params_df <- getModelParamsDf(temp_df)
boot_df <- bootstrapNullAlternativeModelFast(
  temp_df, params_df = temp_params_df, B = 20)  


nkurzaw/TPP2D documentation built on May 9, 2023, 10:04 a.m.