ProtE (version 1.0.3)

pd_multi: Proteome Discoverer (PD) multiple-files' proteomic analysis

Description

Takes as input Proteomics Data (output of PD) in the format of a single file per sample and creates a master table with the Protein names and Abundance values. Then it performs exploratory data analysis, providing different options for data manipulation (normalization, filtering based on the missing values and imputation) It then proceeds to perform statistical analysis, while creating exploratory plots such as relative log expression boxplots and violin plots, heatmaps and PCA plots.

Usage

pd_multi(
  ...,
  imputation = FALSE,
  global_filtering = TRUE,
  independent = TRUE,
  filtering_value = 50,
  bugs = 0,
  normalization = FALSE,
  parametric = FALSE,
  significance = "p",
  description = FALSE
)

Value

Excel files with the proteomic values that are optionally processed, via normalization, imputation and filtering of proteins with a selected percentage of missing values. The result of the processing is visualized with an Protein Rank Abundance plot. PCA plots for all groups and for just their significant correlations are created. Furthermore violin and boxplots for the proteins of each sample is created and a heatmap for the significant proteins.

Arguments

...

The specific path to the folder where the samples from each group are located. They are passed as unnamed arguments via "...". Attention: Ensure paths use '/' as a directory separator.

imputation

Imputes all remaining missing values. Available methods: "LOD" for assigning the dataset's Limit Of Detection (lowest protein intensity identified), "LOD/2", "Gaussian_LOD" for selecting random values from the normal distribution around LOD with sd= 0.2*LOD, "zeros" for simply assigning 0 to MVs, mean" for replacing missing values with the mean of each protein across the entire dataset, "kNN" for a k-nearest neighbors imputation using 5 neighbors (from the package VIM) and "missRanger" for a random forest based imputation using predictive mean matching (from the package missRanger). By default it is set to FALSE (skips imputation).

global_filtering

TRUE/FALSE. If TRUE, the per-protein percentage of missing values will be calculated across the entire dataset. If FALSE, it will be calculated separately for each group, allowing proteins to remain in the analysis if they meet the criteria within any group. By default it is set to TRUE.

independent

TRUE/FALSE If TRUE, the samples come from different populations, if FALSE they come from the same population (Dependent samples). By default, it is set to TRUE. If set to FALSE, the numbers given in the samples_per_group param must be equal to each other.

filtering_value

The maximum allowable percentage of missing values for a protein. Proteins with missing values exceeding this percentage will be excluded from the analysis. By default it is set to 50.

bugs

Either 0 to treat Proteome Discoverer bugs as Zeros (0) or "average" to convert them into the average of the protein between the samples. By default, it is set to 0. Bugs are referred to to the proteins with empty values inside a single-file analysis

normalization

The specific method for normalizing the data.By default it is set to FALSE. Options are FALSE for no normalization of the data, "log2" for a simple log2 transformation, "Quantile" for a quantiles based normalization and "Cyclic_Loess" for a Cyclic Loess normalization of the log2 data, "median" for a median one, "TIC" for Total Ion Current normalization, "VSN" for Variance Stabilizing Normalization and "PPM" for Parts per Million transformation of the data.

parametric

TRUE/FALSE. Specifies the statistical tests that will be taken into account for creating the PCA plots and heatmap. By default it is set to FALSE (non-parametric).

significance

"p" or "BH" Specifies which of the p-values (nominal vs BH adjusted for multiple hypothesis) will be taken into account for creating the PCA plots and the heatmap. By default it is set to "p" (nominal p-value).

description

TRUE/FALSE. If TRUE, establishes connection to the Uniprot database (via the Uniprot.ws package) and adds the "Description" annotation in the data. This option requires protein Accession IDs and is thus applicable only to the pg.matrix file. It requires also internet access. By default it is set to FALSE (No description fetching).

Examples

#Example of running the function with paths for three groups.


T1_path <- system.file("extdata", "PDexports(multiple_files)",
 "T1_BLCA", package = "ProtE")
T2_path <- system.file("extdata", "PDexports(multiple_files)",
"T2_BLCA", package = "ProtE")

temp_dir1 <- file.path(tempdir(), "T1_path")
temp_dir2 <- file.path(tempdir(), "T2_path")

dir.create(temp_dir1, recursive = TRUE, showWarnings = FALSE)
dir.create(temp_dir2, recursive = TRUE, showWarnings = FALSE)



excel_files1 <- list.files(T1_path, pattern = "\\.xlsx$", full.names = TRUE)
excel_files2 <- list.files(T2_path, pattern = "\\.xlsx$", full.names = TRUE)

file.copy(excel_files1, temp_dir1)
file.copy(excel_files2, temp_dir2)

pd_multi(temp_dir1, temp_dir2,
         normalization = FALSE,
         global_filtering = TRUE,  imputation = FALSE,
         independent = TRUE)