Title:	Multi-Omic Differentially Expressed Gene-Gene Pairs
Version:	1.1.0
Maintainer:	Elisabetta Sciacca <e.sciacca@qmul.ac.uk>
Description:	Performs multi-omic differential network analysis by revealing differential interactions between molecular entities (genes, proteins, transcription factors, or other biomolecules) across the omic datasets provided. For each omic dataset, a differential network is constructed where links represent statistically significant differential interactions between entities. These networks are then integrated into a comprehensive visualization using distinct colors to distinguish interactions from different omic layers. This unified display allows interactive exploration of cross-omic patterns, such as differential interactions present at both transcript and protein levels. For each link, users can access differential statistical significance metrics (p values or adjusted p values, calculated via robust or traditional linear regression with interaction term) and differential regression plots. The methods implemented in this package are described in Sciacca et al. (2023) <doi:10.1093/bioinformatics/btad192>.
License:	GPL-3
Encoding:	UTF-8
LazyData:	true
LazyDataCompression:	gzip
RoxygenNote:	7.3.2
Language:	en-gb
URL:	https://github.com/elisabettasciacca/multiDEGGs/, https://elisabettasciacca.github.io/multiDEGGs/
BugReports:	https://github.com/elisabettasciacca/multiDEGGs/issues
Suggests:	kernlab, nestedcv, pls, qvalue, randomForest, ranger, testthat (≥ 3.0.0)
Imports:	DT, grDevices, graphics, knitr, MASS, magrittr, methods, parallel, pbapply, pbmcapply, rmarkdown, sfsmisc, shiny, shinydashboard, stats, utils, visNetwork
Depends:	R (≥ 4.4.0)
VignetteBuilder:	knitr
Config/testthat/edition:	3
NeedsCompilation:	no
Packaged:	2025-07-29 16:12:14 UTC; elisabetta
Author:	Elisabetta Sciacca [aut, cre, cph], Myles Lewis [ctb]
Repository:	CRAN
Date/Publication:	2025-07-29 16:30:02 UTC

Predict method for multiDEGGs_filter objects

Description

This function generates predictions by creating a dataset with single and combined predictors based on the filtering results of a multiDEGGs_filter model.

Usage

.predict_multiDEGGs(
  object,
  newdata,
  interaction.type = "ratio",
  sep = ":",
  ...
)

Arguments

Details

The function processes the filtering results in two steps:

1. Selects single predictors from newdata based on variables listed in object$keep

2. Adds combined predictors from paired variables in object$pairs

Value

A data frame containing:

• Single predictors (if any are specified in object$keep)

• Combined predictors based on variable pairs and interaction type

Interactive visualisation of differential networks

Description

Explore differential networks and interactively select regression and box plots

Usage

View_diffNetworks(deggs_object, legend.arrow.width = 0.35, stepY_legend = 55)

Arguments

Value

a shiny interface showing networks with selectable nodes and links

Examples

data("synthetic_metadata")
data("synthetic_rnaseqData")
data("synthetic_proteomicData")
data("synthetic_OlinkData")
assayData_list <- list("RNAseq" = synthetic_rnaseqData,
                       "Proteomics" = synthetic_proteomicData,
                       "Olink" = synthetic_OlinkData)
deggs_object <- get_diffNetworks(assayData = assayData_list,
                                 metadata = synthetic_metadata,
                                 category_variable = "response",
                                 regression_method = "lm",
                                 verbose = FALSE,
                                 show_progressBar = FALSE,
                                 cores = 1)
# the below function runs a shiny app, so can't be run during R CMD check                                  
if(interactive()){
View_diffNetworks(deggs_object)  
}

Calculate the p values for specific category network samples

Description

Calculate the p values for specific category network samples

Usage

calc_pvalues_network(
  assayData,
  metadata,
  padj_method,
  categories_length,
  regression_method = "lm",
  category_network
)

Arguments

Value

a list of p values

Compute interaction p values for a single percentile value

Description

Compute interaction p values for a single percentile value

Usage

calc_pvalues_percentile(
  assayData,
  metadata,
  categories_length,
  category_median_list,
  padj_method,
  percentile,
  contrasts,
  regression_method,
  edges,
  sig_edges_count
)

Arguments

Value

The list of float numbers of the significant pvalues for a single percentile

cat_parallel (from nestedcv)

Description

Prints using shell echo from inside mclapply when run in Rstudio

Usage

cat_parallel(...)

Arguments

Generate multi-omic differential networks

Description

Generate a multi-layer differential network with interaction p values

Usage

get_diffNetworks(
  assayData,
  metadata,
  category_variable = NULL,
  regression_method = "lm",
  category_subset = NULL,
  network = NULL,
  percentile_vector = seq(0.35, 0.98, by = 0.05),
  padj_method = "bonferroni",
  show_progressBar = TRUE,
  verbose = TRUE,
  cores = parallel::detectCores()/2
)

Arguments

Value

a deggs object containing differential networks incorporating p values or adjusted p values for each link.

Examples

# Single omic analysis:
data("synthetic_metadata")
data("synthetic_rnaseqData")
deggs_object <- get_diffNetworks(assayData = synthetic_rnaseqData,
                                 metadata = synthetic_metadata,
                                 category_variable = "response",
                                 regression_method = "lm",
                                 padj_method = "bonferroni",
                                 verbose = FALSE,
                                 show_progressBar = FALSE,
                                 cores = 1)
 
# multi-omic analysis: 
data("synthetic_metadata")
data("synthetic_rnaseqData")
data("synthetic_proteomicData")
data("synthetic_OlinkData")
assayData_list <- list("RNAseq" = synthetic_rnaseqData,
                       "Proteomics" = synthetic_proteomicData,
                       "Olink" = synthetic_OlinkData)
deggs_object <- get_diffNetworks(assayData = assayData_list,
                                 metadata = synthetic_metadata,
                                 category_variable = "response",
                                 regression_method = "lm",
                                 padj_method = "bonferroni",
                                 verbose = FALSE,
                                 show_progressBar = FALSE,
                                 cores = 1)
                                 
# to use only certain categories for comparison: 
# let's randomly add another level of response to the example metadata
synthetic_metadata$response <- as.character(synthetic_metadata$response)
indices <- sample(1:nrow(synthetic_metadata), 20, replace = FALSE) 
synthetic_metadata$response[indices] <- "Moderate response"
deggs_object <- get_diffNetworks(assayData = assayData_list,
                                 metadata = synthetic_metadata,
                                 category_variable = "response",
                                 category_subset = c("Responder", 
                                                     "Non_responder"),
                                 regression_method = "lm",
                                 verbose = FALSE,
                                 show_progressBar = FALSE,
                                 cores = 1)
                                 
# to be more generous on the targets to be excluded, and lower the expression 
# level threshold to the 25th percentile (or lower): 
deggs_object <- get_diffNetworks(assayData = assayData_list,
                                 metadata = synthetic_metadata,
                                 category_variable = "response",
                                 category_subset = c("Responder", 
                                                     "Non_responder"),
                                 regression_method = "lm",
                                 percentile_vector = seq(0.25, 0.98, by = 0.05),
                                 verbose = FALSE,
                                 show_progressBar = FALSE,
                                 cores = 1)                                  

Generate differential networks for single omic analysis

Description

Generate differential networks for single omic analysis

Usage

get_diffNetworks_singleOmic(
  assayData,
  assayDataName,
  metadata,
  regression_method,
  network,
  percentile_vector,
  padj_method,
  show_progressBar,
  verbose,
  cores
)

Arguments

Value

a list of differential networks, one per category

Get a table of all significant interactions across categories

Description

Get a table of all significant interactions across categories

Usage

get_multiOmics_diffNetworks(deggs_object, sig_threshold = 0.05)

Arguments

Value

a list of multilayer networks (as edge tables), one per category.

Examples

data("synthetic_metadata")
data("synthetic_rnaseqData")
data("synthetic_proteomicData")
data("synthetic_OlinkData")
assayData_list <- list("RNAseq" = synthetic_rnaseqData,
                       "Proteomics" = synthetic_proteomicData,
                       "Olink" = synthetic_OlinkData)
deggs_object <- get_diffNetworks(assayData = assayData_list,
                                 metadata = synthetic_metadata,
                                 category_variable = "response",
                                 verbose = FALSE,
                                 show_progressBar = FALSE,
                                 cores = 2)
get_multiOmics_diffNetworks(deggs_object, sig_threshold = 0.05)

Get a table of all the significant interactions across categories

Description

Get a table of all the significant interactions across categories

Usage

get_sig_deggs(deggs_object, assayDataName = 1, sig_threshold = 0.05)

Arguments

Value

a data.frame listing all the significant differential interactions found across categories for that particular omic data. This list can also be used to substitute or integrate feature selection in machine learning models for the prediction of the categories (see vignette).

Examples

data("synthetic_metadata")
data("synthetic_rnaseqData")
deggs_object <- get_diffNetworks(assayData = synthetic_rnaseqData,
                                 metadata = synthetic_metadata,
                                 category_variable = "response",
                                 verbose = FALSE,
                                 show_progressBar = FALSE,
                                 cores = 2)
get_sig_deggs(deggs_object, sig_threshold = 0.05) 

Combined multiDEGGs filter

Description

This function can be passed to the modifyX parameter of nestcv.train or nestcv.glmnet to use one of the available statistical filters (t-test, wilcoxon, etc.) in combination with multiDEGGs. Single predictors will be selected by the selected statistical filter an paired predictors will be added by multiDEGGs.

Usage

multiDEGGs_combined_filter(
  y,
  x,
  filter_method = "ttest",
  nfilter,
  dynamic_nfilter = TRUE,
  keep_single_genes = FALSE,
  ...
)

Arguments

Details

The function operates in two modes:

Dynamic Filtering (dynamic_nfilter = TRUE):

• Selects nfilter single genes using the specified statistical method

• Finds all significant gene pairs using multiDEGGs

• Total predictors = nfilter single genes + number of significant pairs

  • If keep_single_genes = TRUE, also includes single genes obtained from pairs found by multiDEGGs

Balanced Selection (dynamic_nfilter = FALSE):

• Allocates approximately half of nfilter to gene pairs

• Remaining slots filled with single genes from the statistical filter

• If fewer pairs are found than allocated, compensates by selecting more single genes

• Ensures consistent total number of predictors across outer folds

The statistical filtering methods include:

• "ttest": Two-sample t-test for differential expression

• "wilcoxon": Wilcoxon rank-sum test

• "ranger": Random Forest variable importance

• "glmnet": Elastic net regularization

• "pls": Partial Least Squares variable importance

Value

An object of class "multiDEGGs_filter" containing:

Examples

library(nestedcv)
data("synthetic_metadata")
data("synthetic_rnaseqData")

# fit a regularized linear model
# note that nfilter, n_outer_folds, n_inner_folds are set low to keep the
# example lightweight. Adjust these values as needed for your use case.
## Not run: 
fit.glmnet <- nestedcv::nestcv.glmnet(
y = as.numeric(synthetic_metadata$response),
x =  t(synthetic_rnaseqData),
modifyX = "multiDEGGs_combined_filter",
modifyX_options = list(filter_method = "ttest", 
                       nfilter = 5,
                       dynamic_nfilter = TRUE, 
                       keep_single_genes = FALSE),
modifyX_useY = TRUE,
n_outer_folds = 4,
n_inner_folds = 4)

summary(fit.glmnet)

## End(Not run)

# fit a random forest model
# NOTE: nfilter, n_outer_folds, n_inner_folds are set low to keep the
# example lightweight. Adjust these values as needed for your use case.
fit.rf <- nestedcv::nestcv.train(
  y = synthetic_metadata$response,
  x = t(synthetic_rnaseqData),
  method = "rf",
  modifyX = "multiDEGGs_combined_filter",
  modifyX_options = list(filter_method = "ttest", 
                         nfilter = 5,
                         dynamic_nfilter = TRUE, 
                         keep_single_genes = FALSE),
  modifyX_useY = TRUE,
  n_outer_folds = 2,
  n_inner_folds = 2
)

fit.rf$summary

multiDEGGs_filter

Description

Function to be passed to the modifyX parameter of nestcv.train or nestcv.glmnet to allow nested feature selection and augmentation via differential network analysis with multiDEGGs.

Usage

multiDEGGs_filter(y, x, keep_single_genes = FALSE, nfilter = NULL)

Arguments

Value

a list containing two types of predictors:

• single predictors (stored in the 'keep_DEGGs' vector)

• paired predictors (stored in the 'pairs' dataframe) Note that nfilter limits the maximum number of engineered features returned, however this number might be lower and will depend on how many significantly different interactions will be found in each fold by multiDEGGs. If no significantly different interactions are found the function will print a '0' and switch to t-test for that fold.

Examples

library(nestedcv)
data("synthetic_metadata")
data("synthetic_rnaseqData")

# fit a regularized linear model
# Note that nfilter, n_outer_folds, n_inner_folds are set low to keep the
# example lightweight. Adjust these values as needed for your use case.
## Not run: 
fit.glmnet <- nestcv.glmnet(
  y = as.numeric(synthetic_metadata$response),
  x =  t(synthetic_rnaseqData),
  modifyX = "multiDEGGs_filter",
  modifyX_options = list(keep_single_genes = FALSE,
                         nfilter = 5),
  modifyX_useY = TRUE,
  n_outer_folds = 4,
  n_inner_folds = 4)

summary(fit.glmnet)

## End(Not run)

# fit a random forest model:
# note that nfilter, n_outer_folds, n_inner_folds are set low to keep the
# example lightweight. Adjust these values as needed for your use case.
fit.rf <- nestcv.train(
  y = synthetic_metadata$response,
  x = t(synthetic_rnaseqData),
  method = "rf",
  modifyX = "multiDEGGs_filter",
  modifyX_options = list(keep_single_genes = FALSE,
                         nfilter = 5),
  modifyX_useY = TRUE,
  n_outer_folds = 2,
  n_inner_folds = 2
)

fit.rf$summary

Internal function for colors

Description

This function return a color palette with the number of colors specified by n

Usage

my_palette(n)

Arguments

Value

a vector with colors

Boxplots of single nodes (genes,proteins, etc.)

Description

This function is for internal use of View_diffnetworks

Usage

node_boxplot(gene, assayDataName = 1, deggs_object)

Arguments

Value

the boxplot

Plot differential regressions for a link

Description

Plot differential regressions for any target-target pair in an omic dataset

Usage

plot_regressions(
  deggs_object,
  assayDataName = 1,
  gene_A,
  gene_B,
  title = NULL,
  legend_position = "topright"
)

Arguments

Value

base graphics plot showing differential regressions across categories. The p value of the interaction term of gene A ~ gene B \* category is reported on top.

Examples

data("synthetic_metadata")
data("synthetic_rnaseqData")
data("synthetic_proteomicData")
data("synthetic_OlinkData")
assayData_list <- list("RNAseq" = synthetic_rnaseqData,
                       "Proteomics" = synthetic_proteomicData,
                       "Olink" = synthetic_OlinkData)
deggs_object <- get_diffNetworks(assayData = assayData_list,
                                 metadata = synthetic_metadata,
                                 category_variable = "response",
                                 regression_method = "lm",
                                 padj_method = "bonferroni",
                                 verbose = FALSE,
                                 show_progressBar = FALSE,
                                 cores = 1)
plot_regressions(deggs_object,
                 assayDataName = "RNAseq",
                 gene_A = "MTOR", 
                 gene_B = "AKT2",
                 legend_position = "bottomright")

Wrapper of .predict_multiDEGGs for multiDEGGs_filter()

Description

This function generates predictions by creating a dataset with single and combined predictors based on the filtering results of a multiDEGGs_filter model.

Usage

## S3 method for class 'multiDEGGs_filter'
predict(object, newdata, interaction.type = "ratio", sep = ":", ...)

Arguments

Details

The function processes the filtering results in two steps:

1. Selects single predictors from newdata based on variables listed in object$keep

2. Adds combined predictors from paired variables in object$pairs

Value

A data frame containing:

• Single predictors (if any are specified in object$keep)

• Combined predictors based on variable pairs and interaction type

Wrapper of .predict_multiDEGGs for multiDEGGs_filter_combined()

Description

This function generates predictions by creating a dataset with single and combined predictors based on the filtering results of a multiDEGGs_filter model.

Usage

## S3 method for class 'multiDEGGs_filter_combined'
predict(object, newdata, interaction.type = "ratio", sep = ":", ...)

Arguments

Details

The function processes the filtering results in two steps:

1. Selects single predictors from newdata based on variables listed in object$keep

2. Adds combined predictors from paired variables in object$pairs

Value

A data frame containing:

• Single predictors (if any are specified in object$keep)

• Combined predictors based on variable pairs and interaction type

Synthetic RNA-seq count data

Description

Synthetic RNA-seq data after log2 normalisation

Format

A data frame with xx rows (proteins) xx columns (patients IDs).

Synthetic clinical data

Description

A dataset containing sample clinical data for 100 patients with 40% response rate

Format

A data frame with 100 rows and 4 columns (IDs are in rownames):

patient_id

IDs matching the IDs used in the colnames of the assay data matrix/matrices.

age

A column to simulate age of patients. Not used.

gender

A column to simulate gender of patients. Not used.

response

The response outcome, to be used for differential analysis

Synthetic RNA-seq count data

Description

Synthetic RNA-seq data after log2 normalisation

Format

A data frame with xx rows (proteins) xx columns (patients IDs).

Synthetic RNA-seq count data

Description

Synthetic RNA-seq data after log2 normalisation

Format

A data frame with xx rows (genes) xx columns (patients IDs, matching the metadata rownames).

Tidying up of metadata. Samples belonging to undesidered categories (if specified) will be removed as well as categories with less than five samples, and NAs.

Description

Tidying up of metadata. Samples belonging to undesidered categories (if specified) will be removed as well as categories with less than five samples, and NAs.

Usage

tidy_metadata(
  category_subset = NULL,
  metadata,
  category_variable = NULL,
  verbose = FALSE
)

Arguments

Value

a tidy named factor vector of sample annotations.