Movatterモバイル変換

Type:	Package
Title:	Proteomics Data Analysis and Modeling Tools
Version:	0.2.2
Description:	A comprehensive, user-friendly package for label-free proteomics data analysis and machine learning-based modeling. Data generated from 'MaxQuant' can be easily used to conduct differential expression analysis, build predictive models with top protein candidates, and assess model performance. promor includes a suite of tools for quality control, visualization, missing data imputation (Lazar et. al. (2016) <doi:10.1021/acs.jproteome.5b00981>), differential expression analysis (Ritchie et. al. (2015) <doi:10.1093/nar/gkv007>), and machine learning-based modeling (Kuhn (2008) <doi:10.18637/jss.v028.i05>).
License:	LGPL-2.1 |LGPL-3 [expanded from: LGPL (≥ 2.1)]
Encoding:	UTF-8
Language:	en-US
RoxygenNote:	7.3.3
VignetteBuilder:	knitr
Suggests:	covr, knitr, rmarkdown, testthat (≥ 3.0.0)
Depends:	R (≥ 3.5.0)
URL:	https://github.com/caranathunge/promor,https://caranathunge.github.io/promor/
Imports:	reshape2, ggplot2, ggrepel, gridExtra, limma, statmod,pcaMethods, VIM, missForest, caret, kernlab, xgboost,naivebayes, viridis, pROC
LazyData:	true
Config/testthat/edition:	3
BugReports:	https://github.com/caranathunge/promor/issues
NeedsCompilation:	no
Packaged:	2025-11-11 16:52:29 UTC; caran
Author:	Chathurani Ranathunge [aut, cre, cph]
Maintainer:	Chathurani Ranathunge <caranathunge86@gmail.com>
Repository:	CRAN
Date/Publication:	2025-11-11 22:20:02 UTC

Compute average intensity

Description

This function computes average intensities acrosstechnical replicates for each sample.

Usage

aver_techreps(raw_df)

Arguments

Details

aver_techreps assumes that column names in the data framefollow the "Group_UniqueSampleID_TechnicalReplicate" notation. (Usehead(raw_df) to see the structure of theraw_df object.)

Value

Araw_df object of averaged intensities.

Author(s)

Chathurani Ranathunge

See Also

• avearrays fromlimmapackage.

• create_df

Examples

## Use a data set containing technical replicates to create a raw_df objectraw_df <- create_df(prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg2.txt",exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed2.txt",tech_reps = TRUE)# Compute average intensities across technical replicates.rawdf_ave <- aver_techreps(raw_df)

Correlation between technical replicates

Description

This function generates scatter plots to visualize thecorrelation between a given pair of technical replicates (Eg: 1 vs 2)for each sample.

Usage

corr_plot(  raw_df,  rep_1,  rep_2,  save = FALSE,  file_type = "pdf",  palette = "viridis",  text_size = 5,  n_row = 4,  n_col = 4,  dpi = 80,  file_path = NULL)

Arguments

Details

• Given a data frame of log-transformed intensities(araw_df object) and a pair of numbers referring to the technicalreplicates,corr_plot produces a list of scatter plots showingcorrelation between the given pair of technical replicates for all thesamples provided in the data frame.

• Note:n_row *n_col should be equal to the number ofsamples to display in a single page.

Value

A list ofggplot2 plot objects.

Author(s)

Chathurani Ranathunge

See Also

create_df

Examples

## Use a data set containing technical replicates to create a raw_df objectraw_df <- create_df(prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg2.txt",exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed2.txt",tech_reps = TRUE)## Compare technical replicates 1 vs. 2 for all samplescorr_plot(raw_df, rep_1 = 1, rep_2 = 2)

Suvarna et al 2021 LFQ data (fit object)

Description

An object of class "MArrayLM" from running find_dep on covid_norm_df

Usage

data(covid_fit_df)

Format

An object of class "MArrayLM"

References

https://www.frontiersin.org/articles/10.3389/fphys.2021.652799/full#h3

Suvarna et al 2021 LFQ data (normalized)

Description

A dataframe containing normalized LFQ protein intensity data for 230proteins in 35 samples (a subset of the original data set)

Usage

data(covid_norm_df)

Format

A data frame with 230 rows (proteins) and 35 columns (samples)

References

https://www.frontiersin.org/articles/10.3389/fphys.2021.652799/full#h3

Create a data frame of protein intensities

Description

This function creates a data frame of protein intensities

Usage

create_df(  prot_groups,  exp_design,  input_type = "MaxQuant",  data_type = "LFQ",  filter_na = TRUE,  filter_prot = TRUE,  uniq_pep = 2,  tech_reps = FALSE,  zero_na = TRUE,  log_tr = TRUE,  base = 2)

Arguments

Details

• This function first reads in the proteinGroups.txt fileproduced by MaxQuant or a standard input file containing a quantitativematrix where the proteins or protein groups are indicated by rows and thesamples by columns.

• It then reads in the expDesign.txt file provided asexp_design and extracts relevant information from it to add to thedata frame. an example of the expDesign.txt is provided here:https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt.

• First, empty rows and columns are removed from the data frame.

• Next, if a proteinGroups.txt file is used, it filters out reverseproteins, proteins that were only identified by site, and potentialcontaminants.Then it removes proteins identified with less thanthe number of unique peptides indicated byuniq_pep from thedata frame.

• Next, it extracts the intensity columns indicated bydata typeand the selected protein rows from the data frame.

• Converts missing values (zeros) to NAs.

• Finally, the function log transforms the intensity values.

Value

Araw_df object which is a data frame containing proteinintensities. Proteins or protein groups are indicated by rows and samplesby columns.

Author(s)

Chathurani Ranathunge

Examples

### Using a proteinGroups.txt file produced by MaxQuant as input.## Generate a raw_df object with default settings. No technical replicates.raw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt",  input_type = "MaxQuant")## Data containing technical replicatesraw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg2.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed2.txt",  input_type = "MaxQuant",  tech_reps = TRUE)## Alter the number of unique peptides needed to retain a proteinraw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt",  input_type = "MaxQuant",  uniq_pep = 1)## Use "iBAQ" values instead of "LFQ" valuesraw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt",  input_type = "MaxQuant",  data_type = "iBAQ")### Using a universal standard input file instead of MaxQuant output.raw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/st.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt",  input_type = "standard")

Cox et al 2014 LFQ data (fit object)

Description

An object of class "MArrayLM" from running find_dep on ecoli_norm_df

Usage

data(ecoli_fit_df)

Format

An object of class "MArrayLM"

References

https://europepmc.org/article/MED/24942700#id609082

Cox et al 2014 LFQ data (normalized)

Description

A dataframe containing normalized LFQ protein intensity data for 4360proteins in 6 samples

Usage

data(ecoli_norm_df)

Format

A data frame with 4360 rows (proteins) and 6 columns (samples)

References

https://europepmc.org/article/MED/24942700#id609082

Visualize feature (protein) variation among conditions

Description

This function visualizes protein intensity differences amongconditions (classes) using box plots or density distribution plots.

Usage

feature_plot(  model_df,  type = "box",  text_size = 10,  palette = "viridis",  n_row,  n_col,  save = FALSE,  file_path = NULL,  file_name = "Feature_plot",  file_type = "pdf",  dpi = 80,  plot_width = 7,  plot_height = 7)

Arguments

Details

This function visualizes condition-wise differences in proteinintensity using boxplots and/or density plots.

Value

Aggplot2 object

Author(s)

Chathurani Ranathunge

See Also

• pre_process,rem_feature

Examples

## Create a model_df object with default settings.covid_model_df <- pre_process(covid_fit_df, covid_norm_df)## Feature variation - box plotsfeature_plot(covid_model_df, type = "box", n_row = 4, n_col = 2)## Density plotsfeature_plot(covid_model_df, type = "density")## Change color palettefeature_plot(covid_model_df, type = "density", n_row = 4, n_col = 2, palette = "rocket")

Filter proteins by group level missing data

Description

This function filters out proteins based on missing dataat the group level.

Usage

filterbygroup_na(raw_df, set_na = 0.34, filter_condition = "either")

Arguments

Details

• This function firstextracts group or condition information from theraw_df object andassigns samples to their groups.

• Iffilter_condition = "each", it then removes proteins (rows)from the data frame if the proportion of NAs ineach group exceeds thethreshold indicated byset_na (default is 0.34). This option ismore lenient in comparison tofilter_condition = "either", whereproteins that exceeds the missing data threshold ineither group getsremoved from the data frame.

Value

Araw_df object.

Author(s)

Chathurani Ranathunge

See Also

create_df

Examples

# Generate a raw_df object with default settings. No technical replicates.raw_df <- create_df(prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt")## Remove proteins that exceed 34% NAs in either group (default)rawdf_filt1 <- filterbygroup_na(raw_df)## Remove proteins that exceed 34% NAs in each grouprawdf_filt2 <- filterbygroup_na(raw_df, filter_condition = "each")## Proportion of samples with NAs allowed in each group = 0.5rawdf_filt3 <- filterbygroup_na(raw_df, set_na = 0.5, filter_condition = "each")

Identify differentially expressed proteins between groups

Description

This function performs differential expression analysison protein intensity data with limma.

Usage

find_dep(  df,  save_output = FALSE,  save_tophits = FALSE,  file_path = NULL,  adj_method = "BH",  cutoff = 0.05,  lfc = 1,  n_top = 20)

Arguments

Details

• It is important that the data is first log-transformed, ideally,imputed, and normalized before performing differential expression analysis.

• save_output saves the complete results table from thedifferential expression analysis.

• save_tophits first subsets the results to those with absolutelog fold change of more than 1, performs multiple correction withthe method specified inadj_method and outputs the topn_topresults based on lowest p-value and adjusted p-value.

• If the number of hits with absolute log fold change of more than 1 isless thann_top,find_dep prints only those withlog-fold change > 1 to "TopHits.txt".

• If thefile_path is not specified, text files will be saved ina temporary directory.

Value

Afit_df object, which is similar to alimmafit object.

Author(s)

Chathurani Ranathunge

References

Ritchie, Matthew E., et al. "limma powers differential expressionanalyses for RNA-sequencing and microarray studies." Nucleic acids research43.7 (2015): e47-e47.

See Also

• normalize_data

• lmFit,eBayes,topTable, andwrite.fit functions from thelimma package.

Examples

## Perform differential expression analysis using default settingsfit_df1 <- find_dep(ecoli_norm_df)## Change p-value and adjusted p-value cutofffit_df2 <- find_dep(ecoli_norm_df, cutoff = 0.1)

Heatmap of differentially expressed proteins

Description

This function generates a heatmap to visualize differentiallyexpressed proteins between groups

Usage

heatmap_de(  fit_df,  df,  adj_method = "BH",  cutoff = 0.05,  lfc = 1,  sig = "adjP",  n_top = 20,  palette = "viridis",  text_size = 10,  save = FALSE,  file_path = NULL,  file_name = "HeatmapDE",  file_type = "pdf",  dpi = 80,  plot_height = 7,  plot_width = 7)

Arguments

Details

By default the tiles in the heatmap are reordered by intensity valuesalong both axes (x axis = samples, y axis = proteins).

Value

Aggplot2 plot object.

Author(s)

Chathurani Ranathunge

See Also

• find_dep

• topTable andlmFit functions from thelimma package.

Examples

## Build a heatmap of differentially expressed proteins using the provided## example fit_df and norm_df data objectsheatmap_de(covid_fit_df, covid_norm_df)## Create a heatmap with P-value of 0.05 and log fold change of 1 as## significance criteria.heatmap_de(covid_fit_df, covid_norm_df, cutoff = 0.05, sig = "P")## Visualize the top 30 differentially expressed proteins in the heatmap and## change the color paletteheatmap_de(covid_fit_df, covid_norm_df,  cutoff = 0.05, sig = "P", n_top = 30,  palette = "magma")

Visualize missing data

Description

This function visualizes the patterns of missing valueoccurrence using a heatmap.

Usage

heatmap_na(  raw_df,  protein_range,  sample_range,  reorder_x = FALSE,  reorder_y = FALSE,  x_fun = mean,  y_fun = mean,  palette = "viridis",  label_proteins = FALSE,  text_size = 10,  save = FALSE,  file_type = "pdf",  file_path = NULL,  file_name = "Missing_data_heatmap",  plot_width = 15,  plot_height = 15,  dpi = 80)

Arguments

Details

This function visualizes patterns of missing value occurrence using aheatmap. The user can choose to reorder the axes using the available functions(x_fun,y_fun) to better understand the underlying cause ofmissing data.

Value

Aggplot2 plot object.

Author(s)

Chathurani Ranathunge

See Also

create_df

Examples

## Generate a raw_df object with default settings. No technical replicates.raw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt")## Missing data heatmap with default settings.heatmap_na(raw_df)## Missing data heatmap with x and y axes reordered by the mean (default) of## protein intensity.heatmap_na(raw_df,  reorder_x = TRUE, reorder_y = TRUE)## Missing data heatmap with x and y axes reordered by the sum of## protein intensity.heatmap_na(raw_df,  reorder_x = TRUE, reorder_y = TRUE, x_fun = sum,  y_fun = sum)## Missing data heatmap for a subset of the proteins with x and y axes## reordered by the mean (default) of protein intensity and the y axis## labeled with protein IDs.heatmap_na(raw_df,  protein_range = 1:30,  reorder_x = TRUE, reorder_y = TRUE,  label_proteins = TRUE)

Impute missing values

Description

This function imputes missing values using a user-specifiedimputation method.

Usage

impute_na(  df,  method = "minProb",  tune_sigma = 1,  q = 0.01,  maxiter = 10,  ntree = 20,  n_pcs = 2,  seed = NULL)

Arguments

Details

• Ideally, you should first remove proteins withhigh levels of missing data using thefilterbygroup_na functionbefore runningimpute_na on theraw_df object or thenorm_df object.

• impute_na function imputes missing values using auser-specified imputation method from the available options,minProb,minDet,kNN,RF, andSVD.

• Note: Some imputation methods may require that the data be normalizedprior to imputation.

• Make sure to fix the random number seed withseed for reproducibility

.

Value

Animp_df object, which is a data frame of protein intensitieswith no missing values.

Author(s)

Chathurani Ranathunge

References

Lazar, Cosmin, et al. "Accounting for the multiple natures ofmissing values in label-free quantitative proteomics data sets to compareimputation strategies." Journal of proteome research 15.4 (2016): 1116-1125.

See Also

More information on the available imputation methods can be foundin their respective packages.

• create_df

• ForminProb andminDet methods, seeimputeLCMD package.

• For Random Forest (RF) method, seemissForest.

• ForkNN method, seekNN from theVIM package.

• ForSVD method, seepca from thepcaMethods package.

Examples

## Generate a raw_df object with default settings. No technical replicates.raw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt")## Impute missing values in the data frame using the default minProb## method.imp_df1 <- impute_na(raw_df, seed = 3312)## Using the kNN method.imp_df2 <- impute_na(raw_df, method = "kNN", seed = 3312)## Using the SVD method with n_pcs set to 3.imp_df3 <- impute_na(raw_df, method = "SVD", n_pcs = 3, seed = 3312)## Using the minDet method with q set at 0.001.imp_df4 <- impute_na(raw_df, method = "minDet", q = 0.001, seed = 3312)## Impute a normalized data set using the kNN methodimp_df5 <- impute_na(ecoli_norm_df, method = "kNN")

Visualize the impact of imputation

Description

This function generates density plots to visualize the impact ofmissing data imputation on the data.

Usage

impute_plot(  original,  imputed,  global = TRUE,  text_size = 10,  palette = "viridis",  n_row,  n_col,  save = FALSE,  file_path = NULL,  file_name = "Impute_plot",  file_type = "pdf",  plot_width = 7,  plot_height = 7,  dpi = 80)

Arguments

Details

• Given two data frames, one with missing valuesand the other, an imputed data frame (imp_df object) of the samedata set,impute_plot generates global or sample-wise density plotsto visualize the impact of imputation on the data set.

• Note, when sample-wise option is selected (global = FALSE),n_col andn_row can be used to specify the number of columnsand rows to print the plots.

• If you choose to specifyn_row andn_col, make sure thatn_row *n_col matches the total number of samples in thedata frame.

Value

Aggplot2 plot object.

Author(s)

Chathurani Ranathunge

Examples

## Generate a raw_df object with default settings. No technical replicates.raw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt")## Impute missing values in the data frame using the default minProb## method.imp_df <- impute_na(raw_df)## Visualize the impact of missing data imputation with a global density## plot.impute_plot(original = raw_df, imputed = imp_df)## Make sample-wise density plotsimpute_plot(raw_df, imp_df, global = FALSE)## Print plots in user-specified numbers of rows and columnsimpute_plot(raw_df, imp_df, global = FALSE, n_col = 2, n_row = 3)

Visualize the effect of normalization

Description

This function visualizes the impact of normalization onthe data

Usage

norm_plot(  original,  normalized,  type = "box",  text_size = 10,  palette = "viridis",  save = FALSE,  file_path = NULL,  file_name = "Norm_plot",  file_type = "pdf",  dpi = 80,  plot_width = 10,  plot_height = 7)

Arguments

Details

Given two data frames, one with data prior to normalization(original), and the other, after normalization (normalized),norm_plot generates side-by-side plots to visualize the effect ofnormalization on the protein intensity data.

Value

Aggplot2 plot object.

Author(s)

Chathurani Ranathunge

See Also

• normalize_data

• create_df

• impute_na

Examples

## Generate a raw_df object with default settings. No technical replicates.raw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt")## Impute missing values in the data frame using the default minProb## method.imp_df <- impute_na(raw_df)## Normalize the imp_df object using the default quantile methodnorm_df <- normalize_data(imp_df)## Visualize normalization using box plotsnorm_plot(original = imp_df, normalized = norm_df)## Visualize normalization using density plotsnorm_plot(imp_df, norm_df, type = "density")

Normalize intensity data

Description

This function normalizes data using a user-specifiednormalization method.

Usage

normalize_data(df, method = "quantile")

Arguments

Details

• normalize_data is a wrapper function aroundthenormalizeBetweenArrays function from thelimma package.

• This function normalizesintensity values to achieve consistency among samples.

• It assumes that the intensities in thedata frame have been log-transformed, therefore, it is important to make surethatcreate_df was run withlog_tr = TRUE(default) whencreating theraw_df object.

Value

Anorm_df object, which is a data frame ofnormalized protein intensities.

Author(s)

Chathurani Ranathunge

See Also

• create_df

• impute_na

• SeenormalizeBetweenArrays in the R packagelimma for more information on the different normalization methodsavailable.

Examples

## Generate a raw_df object with default settings. No technical replicates.raw_df <- create_df(  prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",  exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt")## Impute missing values in the data frame using the default minProb## method prioir to normalization.imp_df <- impute_na(raw_df)## Normalize the imp_df object using the default quantile methodnorm_df1 <- normalize_data(imp_df)## Use the cyclicloess methodnorm_df2 <- normalize_data(imp_df, method = "cyclicloess")## Normalize data in the raw_df object prior to imputation.norm_df3 <- normalize_data(raw_df)

Proteins that are only expressed in a given group

Description

This function outputs a list of proteins that are onlyexpressed (present) in one user-specified group while not expressed(completely absent) in another user-specified group.

Usage

onegroup_only(  raw_df,  abs_group,  pres_group,  set_na = 0.34,  save = FALSE,  file_path = NULL)

Arguments

Details

Note:onegroup_only function assumes that column names intheraw_df object provided asdf follow "Group_UniqueSampleID"notation. (Usehead(raw_df) to check the structure of yourraw_df object.)

• Given a pair of groups,onegroup_onlyfunction finds proteins that are only expressed inpres_group whilecompletely absent or not expressed inabs_group.

• A text file containing majority protein IDs will be saved in atemporary directory iffile_path is not specified.

Value

A list of majority protein IDs.

Author(s)

Chathurani Ranathunge

Examples

# Generate a raw_df object with default settings. No technical replicates.raw_df <- create_df(prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg1.txt",exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed1.txt")## Find the proteins only expressed in group L, but absent in group H.onegroup_only(raw_df, abs_group = "H",pres_group = "L")

Model performance plot

Description

This function generates plots to visualize model performance

Usage

performance_plot(  model_list,  type = "box",  text_size = 10,  palette = "viridis",  save = FALSE,  file_path = NULL,  file_name = "Performance_plot",  file_type = "pdf",  plot_width = 7,  plot_height = 7,  dpi = 80)

Arguments

Details

• performance_plot uses resampling results frommodels included in themodel_list to generate plots showing modelperformance.

• The default metrics used for classification based models are "Accuracy"and "Kappa."

• These metric types can be changed by providing additional arguments tothetrain_models function. Seetrain andtrainControl for more information.

Value

Aggplot2 object.

Author(s)

Chathurani Ranathunge

See Also

• train_models

• resamples

• train

• trainControl

Examples

## Create a model_df objectcovid_model_df <- pre_process(covid_fit_df, covid_norm_df)## Split the data frame into training and test data setscovid_split_df <- split_data(covid_model_df)## Fit models based on the default list of machine learning (ML) algorithmscovid_model_list <- train_models(covid_split_df)## Generate box plots to visualize performance of different ML algorithmsperformance_plot(covid_model_list)## Generate dot plotsperformance_plot(covid_model_list, type = "dot")## Change color paletteperformance_plot(covid_model_list, type = "dot", palette = "inferno")

Pre-process protein intensity data for modeling

Description

This function pre-processes protein intensity data fromthe top differentially expressed proteins identified withfind_dep formodeling.

Usage

pre_process(  fit_df,  norm_df,  sig = "adjP",  sig_cutoff = 0.05,  fc = 1,  n_top = 20,  find_highcorr = TRUE,  corr_cutoff = 0.9,  save_corrmatrix = FALSE,  file_path = NULL,  rem_highcorr = TRUE)

Arguments

Details

This function creates a data frame that contains protein intensitiesfor a user-specified number of top differentially expressed proteins.

• Usingfind_highcorr = TRUE, highly correlatedproteins can be identified, and can be removed withrem_highcorr = TRUE.

• Note: Most models will benefit from reducing correlation betweenproteins (predictors or features), therefore we recommend removing thoseproteins at this stage to reduce pairwise-correlation.

• If no or few proteins meet the significance threshold for differentialexpression, you may adjustsig,fc, and/orsig_cutoffaccordingly to obtain more proteins for modeling.

Value

Amodel_df object, which is a data frame of proteinintensities with proteins indicated by columns.

Author(s)

Chathurani Ranathunge

See Also

• find_dep,normalize_data

• caret: findCorrelation

Examples

## Create a model_df object with default settings.covid_model_df1 <- pre_process(fit_df = covid_fit_df, norm_df = covid_norm_df)## Change the correlation cutoff.covid_model_df2 <- pre_process(covid_fit_df, covid_norm_df, corr_cutoff = 0.95)## Change the significance criteria to include more proteinscovid_model_df3 <- pre_process(covid_fit_df, covid_norm_df, sig = "P")## Change the number of top differentially expressed proteins to includecovid_model_df4 <- pre_process(covid_fit_df, covid_norm_df, sig = "P", n_top = 24)

Remove user-specified proteins (features) from a data frame

Description

This function removes user-specified proteins from amodel_dfobject

Usage

rem_feature(model_df, rem_protein)

Arguments

Details

• After visualizing protein intensity variationamong conditions withfeature_plot or after assessing the importanceof each protein in models usingvarimp_plot, you can choose to removespecific proteins (features) from the data frame.

• For example, you canchoose to remove a protein from themodel_df object if the proteindoes not show distinct patterns of variation among conditions. This proteinmay show mostly overlapping distributions in the feature plots.

• Another incidence would be removing a protein that is very low invariable importance in the models built usingtrain_models. You canvisualize variable importance usingvarimp_plot.

Value

Amodel_df object.

Author(s)

Chathurani Ranathunge

See Also

feature_plot,pre_process

Examples

covid_model_df <- pre_process(fit_df = covid_fit_df, norm_df = covid_norm_df)## Remove sp|P22352|GPX3_HUMAN protein from the model_df objectcovid_model_df1 <- rem_feature(covid_model_df, rem_protein = "sp|P22352|GPX3_HUMAN")

Remove user-specified samples

Description

This function removes user-specified samples from thedata frame.

Usage

rem_sample(raw_df, rem)

Arguments

Details

• rem_sample assumes that sample names follow the"Group_UniqueSampleID_TechnicalReplicate" notation (Usehead(raw_df)to see the structure of theraw_df object.)

• If all the technical replicates representing a sample needs to beremoved, provide "Group_UniqueSampleID" asrem.

• If a specific technical replicate needs to be removed in case itshows weak correlation with other technical replicates for example, you canremove that particular technical replicate by providing"Group_UniqueSampleID_TechnicalReplicate" asrem.

Value

Araw_df object.

Author(s)

Chathurani Ranathunge

See Also

corr_plot,create_df

Examples

## Use a data set containing technical replicates to create a raw_df objectraw_df <- create_df(prot_groups = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/pg2.txt",exp_design = "https://raw.githubusercontent.com/caranathunge/promor_example_data/main/ed2.txt",tech_reps = TRUE)# Check the first few rows of the raw_df objecthead(raw_df)## Remove all technical replicates of "WT_4"raw_df1 <- rem_sample(raw_df, "WT_4")## Remove only technical replicate number 2 of "WT_4"raw_df2 <- rem_sample(raw_df, "WT_4_2")

ROC plot

Description

This function generates Receiver Operating Characteristic (ROC)curves to evaluate models

Usage

roc_plot(  probability_list,  split_df,  ...,  multiple_plots = TRUE,  text_size = 10,  palette = "viridis",  save = FALSE,  file_path = NULL,  file_name = "ROC_plot",  file_type = "pdf",  plot_width = 7,  plot_height = 7,  dpi = 80)

Arguments

Details

• roc_plot first uses probabilities generatedduringtest_models to build a ROC object.

• Next, relevant information is extracted from the ROC object toplot the ROC curves.

Value

Aggplot2 object.

Author(s)

Chathurani Ranathunge

See Also

• test_models

• roc

Examples

## Create a model_df objectcovid_model_df <- pre_process(covid_fit_df, covid_norm_df)## Split the data frame into training and test data setscovid_split_df <- split_data(covid_model_df)## Fit models using the default list of machine learning (ML) algorithmscovid_model_list <- train_models(covid_split_df)# Test a list of models on a test data set and output class probabilities,covid_prob_list <- test_models(covid_model_list, covid_split_df, type = "prob")## Plot ROC curves separately for each ML algorithmroc_plot(covid_prob_list, covid_split_df)## Plot all ROC curves in one plotroc_plot(covid_prob_list, covid_split_df, multiple_plots = FALSE)## Change color paletteroc_plot(covid_prob_list, covid_split_df, palette = "plasma")

Split the data frame to create training and test data

Description

This function can be used to create balanced splits of theprotein intensity data in amodel_df object to create training and test data

Usage

split_data(model_df, train_size = 0.8, seed = NULL)

Arguments

Details

This function splits themodel_df object in to training andtest data sets using random sampling while preserving the originalclass distribution of the data. Make sure to fix the random number seed withseed for reproducibility

Value

A list of data frames.

Author(s)

Chathurani Ranathunge

See Also

• pre_process

• createDataPartition

Examples

## Create a model_df objectcovid_model_df <- pre_process(covid_fit_df, covid_norm_df)## Split the data frame into training and test data sets using default settingscovid_split_df1 <- split_data(covid_model_df, seed = 8314)## Split the data frame into training and test data sets with 70% of the## data in training and 30% in test data setscovid_split_df2 <- split_data(covid_model_df, train_size = 0.7, seed = 8314)## Access training data setcovid_split_df1$training## Access test data setcovid_split_df1$test

Test machine learning models on test data

Description

This function can be used to predict test data using modelsgenerated by different machine learning algorithms

Usage

test_models(  model_list,  split_df,  type = "prob",  save_confusionmatrix = FALSE,  file_path = NULL,  ...)

Arguments

Details

• test_models function usesmodels obtained fromtrain_models to predict a given test data set.

• Settingtype = "raw" is required to obtain confusion matrices.

• Settingtype = "prob" (default) will output a list ofprobabilities that can be used to generate ROC curves usingroc_plot.

Value

• probability_list: Iftype = "prob", a list ofdata frames containing class probabilities for each method in themodel_list will be returned.

• prediction_list: Iftype = "raw", a list of factorscontaining class predictions for each method will be returned.

Author(s)

Chathurani Ranathunge

See Also

• split_df

• train_models

• predict

• confusionMatrix

Examples

## Create a model_df objectcovid_model_df <- pre_process(covid_fit_df, covid_norm_df)## Split the data frame into training and test data setscovid_split_df <- split_data(covid_model_df)## Fit models using the default list of machine learning (ML) algorithmscovid_model_list <- train_models(covid_split_df)# Test a list of models on a test data set and output class probabilities,covid_prob_list <- test_models(model_list = covid_model_list, split_df = covid_split_df)## Not run: # Save confusion matrices in the working directory and output class predictionscovid_pred_list <- test_models(  model_list = covid_model_list,  split_df = covid_split_df,  type = "raw",  save_confusionmatrix = TRUE,  file_path = ".")## End(Not run)

Train machine learning models on training data

Description

This function can be used to train models on protein intensitydata using different machine learning algorithms

Usage

train_models(  split_df,  resample_method = "repeatedcv",  resample_iterations = 10,  num_repeats = 3,  algorithm_list,  seed = NULL,  ...)

Arguments

Details

• train_models function can be used to firstdefine the control parameters to be used in training models, calculateresampling-based performance measures for models based on a given set ofmachine-learning algorithms, and output the best model for each algorithm.

• In the event thatalgorithm_list is not provided, a defaultlist of four classification-based machine-learning algorithms will be usedfor building and training models. Defaultalgorithm_list:"svmRadial", "rf", "glm", "xgbLinear, and "naive_bayes."

• Note: Models that fail to build are removed from the output.

• Make sure to fix the random number seed withseed for reproducibility

Value

A list of classtrain for each machine-learning algorithm.Seetrain for more information on accessingdifferent elements of this list.

Author(s)

Chathurani Ranathunge

References

Kuhn, Max. "Building predictive models in R using the caretpackage." Journal of statistical software 28 (2008): 1-26.

See Also

• pre_process

• trainControl

• train

Examples

## Create a model_df objectcovid_model_df <- pre_process(covid_fit_df, covid_norm_df)## Split the data frame into training and test data setscovid_split_df <- split_data(covid_model_df, seed = 8314)## Fit models based on the default list of machine learning (ML) algorithmscovid_model_list1 <- train_models(split_df = covid_split_df, seed = 351)## Fit models using a user-specified list of ML algorithms.covid_model_list2 <- train_models(  covid_split_df,  algorithm_list = c("svmRadial", "glmboost"),  seed = 351)## Change resampling method and resampling iterations.covid_model_list3 <- train_models(  covid_split_df,  resample_method = "cv",  resample_iterations = 50,  seed = 351)

Variable importance plot

Description

This function visualizes variable importance in models

Usage

varimp_plot(  model_list,  ...,  type = "lollipop",  text_size = 10,  palette = "viridis",  n_row,  n_col,  save = FALSE,  file_path = NULL,  file_name = "VarImp_plot",  file_type = "pdf",  dpi = 80,  plot_width = 7,  plot_height = 7)

Arguments

Details

• varimp_plot produces a list of plots showingvariable importance measures calculated from models generated with differentmachine-learning algorithms.

• Note: Variables are ordered by variable importance indescending order, and by default, importance values are scaled to 0 and 100.This can be changed by specifyingscale = FALSE. SeevarImp for more information.

Value

A list ofggplot2 objects.

Author(s)

Chathurani Ranathunge

See Also

• train_models,rem_feature

• varImp

Examples

## Create a model_df objectcovid_model_df <- pre_process(covid_fit_df, covid_norm_df)## Split the data frame into training and test data setscovid_split_df <- split_data(covid_model_df)## Fit models based on the default list of machine learning (ML) algorithmscovid_model_list <- train_models(covid_split_df)## Variable importance - lollipop plotsvarimp_plot(covid_model_list)## Bar plotsvarimp_plot(covid_model_list, type = "bar")## Do not scale variable importance valuesvarimp_plot(covid_model_list, scale = FALSE)## Change color palettevarimp_plot(covid_model_list, palette = "magma")

Volcano plot

Description

This function generates volcano plots to visualizedifferentially expressed proteins between groups.

Usage

volcano_plot(  fit_df,  adj_method = "BH",  sig = "adjP",  cutoff = 0.05,  lfc = 1,  line_fc = TRUE,  line_p = TRUE,  palette = "viridis",  text_size = 10,  label_top = FALSE,  n_top = 10,  save = FALSE,  file_path = NULL,  file_name = "Volcano_plot",  file_type = "pdf",  plot_height = 7,  plot_width = 7,  dpi = 80)

Arguments

Details

• Volcano plots show log-2-fold change on the x-axis,and based on the significance criteria chosen, either -log10(p-value) or-log10(adjusted p-value) on the y-axis.

• volcano_plot requires afit_df object from performingdifferential expression analysis withfind_dep.

• User has the option to choose criteria that denote significance.

Value

Aggplot2 plot object.

Author(s)

Chathurani Ranathunge

See Also

• find_dep

• topTable andlmFit functions from thelimma package.

Examples

## Create a volcano plot with default settings.volcano_plot(ecoli_fit_df)## Change significance criteria and cutoffvolcano_plot(ecoli_fit_df, cutoff = 0.1, sig = "P")## Label top 30 differentially expressed proteins and## change the color palette of the plotvolcano_plot(ecoli_fit_df, label_top = TRUE, n_top = 30, palette = "mako")