| Type: | Package |
| Title: | Network-Guided Penalized Regression (NetGreg) |
| Version: | 0.0.2 |
| Description: | A network-guided penalized regression framework that integrates network characteristics from Gaussian graphical models with partial penalization, accounting for both network structure (hubs and non-hubs) and clinical covariates in high-dimensional omics data, including transcriptomics and proteomics. The full methodological details can be found in our recent preprint by Ahn S and Oh EJ (2025) <doi:10.48550/arXiv.2505.22986>. |
| License: | GPL-3 |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.3.2 |
| Depends: | R (≥ 3.5.0) |
| Imports: | huge, glmnet, dplyr, stats, plsgenomics |
| Suggests: | knitr, rmarkdown, testthat (≥ 3.0.0) |
| NeedsCompilation: | no |
| Packaged: | 2025-05-30 04:40:17 UTC; seungjunahn |
| Author: | Seungjun Ahn 
[cre, aut],
Eun Jeong Oh
[aut] |
| Maintainer: | Seungjun Ahn <seungjun.ahn@mountsinai.org> |
| Repository: | CRAN |
| Date/Publication: | 2025-06-03 09:30:12 UTC |
NetworkGuided
Description
A main function to obtain network-guided penalized regression coefficient estimates.
Usage
NetworkGuided(Y, X, hubs, Z, nfolds = 5)
Arguments
Y |
A continuous outcome variable. |
X |
A data matrix of dimension n x p representing samples (rows) by features (columns). |
hubs |
A vector of hubs idenfitied through identifyHubs function from our package. |
Z |
A matrix of clinical or demographic covariates. |
nfolds |
A user-specified numeric value for k-fold cross-validation. |
Value
A vector of network-guided penalized regression coefficients.
Examples
library(plsgenomics)
data(Colon) ## Data from plsgenomics R package
X = data.frame(Colon$X[,1:100]) ## The first 100 genes
Z = data.frame(Colon$X[,101:102]) ## Two clinical covariates
colnames(Z) = c("Z1", "Z2")
Y = as.vector(Colon$X[,1000]) ## Continuous outcome variable
## Apply identifyHubs():
preNG = identifyHubs(X=X, delta=0.05, tau=5, ebic.gamma = 0.1)
## Explore preNG results:
hubs = preNG$hubs ## Returns the names of the identified hub nodes.
## Use our main NetworkGuided function, to obtain network-guided
## penalized regression coefficient estimates.
NG = NetworkGuided(Y=Y, X=X, hubs=preNG$hubs, Z=Z, nfolds=5)
NG$coef
identifyHubs
Description
A function to identify hub nodes (i.e., genes or proteins) from high-dimensional data using network-based criteria.
Usage
identifyHubs(X, delta, tau, ebic.gamma = 0.1)
Arguments
X |
A data matrix of dimension n x p representing samples (rows) by features (columns). |
delta |
A numeric value indicating the proportion of nodes to considered as hubs in a network. |
tau |
A user-specified cutoff for the number of hubs. |
ebic.gamma |
A numeric value specifying the tuning parameter for the extended Bayesian information criterion (eBIC) used in network estimation. |
Value
A list containing (1) the selected sparse graph structure and model selection results; (2) a data frame of feature names with their associated network characteristics (e.g., degree centrality); and (3) a character vector of top-ranked hub features (e.g., hub genes or proteins).
Examples
library(plsgenomics)
data(Colon) ## Data from plsgenomics R package
X = data.frame(Colon$X[,1:100]) ## The first 100 genes
Z = data.frame(Colon$X[,101:102]) ## Two clinical covariates
colnames(Z) = c("Z1", "Z2")
Y = as.vector(Colon$X[,1000]) ## Continuous outcome variable
## Apply identifyHubs():
preNG = identifyHubs(X=X, delta=0.05, tau=5, ebic.gamma = 0.1)
## Explore preNG results:
## To display the degree centrality for each node,
## sorted from strongest to weakest.
preNG$assoResults
preNG$hubs ## Returns the names of the identified hub nodes.