Understand topology and influence.
Reconstruct networks, quantify centrality, examine associations among centrality measures and use integrative scores to identify influential nodes, hubs and spreaders.
Associations→Network→Centrality→IVI / scores
CRAN
GPL-3
NetworkAnalysis
bulk + single-cell omics
Find what matters.
From networks to biological candidates.
influential unifies network reconstruction, multi-scale centrality, influence ranking and experimental feature prioritization in one R toolkit.
Analyze complex networks with IVI, quantify local hubs and information spread, simulate topology-level influence with SIRIR, or move directly from experimental omics data to ranked drivers, biomarkers and mediators with ExIR.
Multi-scaleLocal, semi-local and global network structure.
IntegrativeCentrality, machine learning and experimental evidence.
FlexibleNetworks, bulk omics and single-cell RNA-seq.
InteractiveBundled and browser-based Shiny interfaces.
IVI · network influence
ExIR · experimental prioritization
Hubness · local power
Spreading · information flow
SIRIR · simulation
From structure to actionable candidatesData → associations → network → influence → prioritized features
Start where you are
Choose the shortest path to your question.
You do not need to learn every function first. Start with the scientific question or data type you already have.
↗I have a network
Rank influential nodes with IVI and inspect complementary centrality dimensions.
↗I have experimental omics dataUse ExIR to classify and prioritize candidate drivers, biomarkers and mediators.
↗I want hubs or spreadersUse Hubness score for local network power or Spreading score for information-diffusion potential.
↗I want simulation-based influenceUse SIRIR to rank nodes by their effect on network topology and information spread.
↗I want to test perturbationsSimulate knockout or up-regulation before experimental validation.
↗I prefer a graphical interfaceLaunch IVI or ExIR through the web portal or the bundled local Shiny apps.
One toolkit, complementary questions
Work from a network, or let experimental data drive the network.
influential supports both classical network-centric analysis and experimental feature prioritization, so the workflow can begin with either relationships or measurements.
Prioritize candidates from omics data.
ExIR combines experimental evidence, machine learning, network reconstruction and influence ranking to classify candidate features into biologically interpretable result classes.
Omics data→Filtering & scoring→Network→Candidates
Core workflows
Build the analysis from association to prioritization.
The package spans the foundational steps required to construct and interrogate networks, then extends them with integrative influence methods and experimental feature ranking.
Association analysis
Calculate Pearson or Spearman relationships, mutual ranks, p-values and adjusted p-values with fcor().
Network reconstruction
Build igraph networks from edge lists, adjacency matrices, incidence matrices or SIF files.
Centrality analysis
Quantify local, semi-local and global structure using complementary centrality measures.
Centrality association
Assess dependence, non-linearity, non-monotonicity and conditional relationships among continuous measures.
Feature prioritization
Integrate experimental evidence and network influence to classify and rank candidates with ExIR.
Featured methods
Six ways to ask what matters most.
Choose a method according to whether your question is about experimental candidates, overall influence, local hubs, spreading potential, simulation or perturbation.
ExIR
Classify and prioritize candidate drivers, biomarkers and mediators from experimental omics data by integrating experimental evidence, machine learning, network reconstruction and influence ranking.
IVI
Combine local, semi-local and global topological information to identify influential nodes while reducing the limitations of relying on any single centrality measure.
Hubness score
Summarize local network power using degree centrality and local H-index to identify nodes that are structurally strong in their surrounding environment.
Spreading score
Prioritize nodes with high potential to spread information by integrating ClusterRank, neighborhood connectivity, betweenness and collective influence.
SIRIR
Combine susceptible–infected–recovered simulations with leave-one-out perturbation to rank nodes according to their effect on topology and information spread.
Computational manipulation
Simulate feature knockout and/or up-regulation and evaluate how those manipulations alter information flow within an association network.
Method selector
Which method should I use?
Use the scientific objective, not the function list, to choose your entry point.
| Your goal | Start with | Typical input | Main output |
|---|---|---|---|
| Find the most influential nodes | IVI | Network / igraph object | Integrated influence score per node |
| Prioritize experimental candidates | ExIR | Omics data + differential/regression evidence | Drivers, biomarkers and mediators |
| Find locally powerful hubs | Hubness score | Network | Local topological power |
| Find strong information spreaders | Spreading score | Network | Spreading potential |
| Measure simulation-based influence | SIRIR | Network | Influence rank from SIR perturbation |
| Evaluate candidate perturbations | comp_manipulate() | ExIR output or independent network | Knockout / up-regulation ranks |
Experimental prioritization
ExIR turns omics measurements into ranked biological candidates.
ExIR accepts experimental data from platforms including transcriptomics, proteomics, bulk RNA-seq and single-cell RNA-seq. It combines multi-level filtration and scoring based on supervised learning, unsupervised learning, network reconstruction and integrated influence ranking.
DriversBiomarkersDE-mediatorsnonDE-mediators
Experimental dataMatrix, sparse matrix, data frame or Seurat object
Differential / regression evidenceEffect sizes, significance and trajectory-related evidence
Scale-aware preparationOptional normalization, pseudo-sampling and conservative feature filtering
→
Integrative scoringSupervised + unsupervised learning
Network reconstructionFeature–feature relationships
Influence rankingNetwork importance contributes to final prioritization
Biological classesDrivers · biomarkers · DE/nonDE mediators
Integrative network influence
IVI captures influence across multiple topological scales.
IVI integrates complementary local, semi-local and global centrality information rather than treating one network property as a universal definition of influence. It can be calculated directly from a graph or from previously calculated component indices.
DegreeClusterRankNeighborhood connectivityLocal H-indexBetweennessCollective influence
LocalDegree + ClusterRank
Semi-localNeighborhood connectivity + local H-index
GlobalBetweenness + collective influence
IVI
Network analysis toolkit
More than a ranking function.
The package provides the upstream and downstream utilities needed to move from measurements to networks, network statistics and interpretable visualizations.
fcor() calculates Pearson/Spearman correlations with optional mutual rank, p-values and adjusted p-values.
Create igraph networks from data frames, adjacency matrices, incidence matrices and SIF files.
Calculate degree, betweenness, neighborhood connectivity, H-index, local H-index, collective influence and ClusterRank.
Evaluate dependence, non-linearity, non-monotonicity and conditional relationships between continuous measures.
Map a centrality measure to node size and color for directed/undirected and weighted/unweighted networks.
Use matrices, sparse matrices, data frames, tibbles or Seurat objects with optional pseudo-sampling for larger datasets.
Quantify the effect of leaving nodes out of the network under susceptible–infected–recovered dynamics.
Simulate knockout and up-regulation to assess effects on information flow before experimental validation.
No-code and local interfaces
Explore IVI and ExIR without writing a full analysis script.
Use the Influential Software Package web portal in a browser, or launch the bundled IVI and ExIR Shiny interfaces locally from R. The apps make core workflows accessible while preserving the scripted package interface for reproducible analysis.
IVI Shiny app
Calculate IVI and visualize networks according to node influence.
influential::runShinyApp("IVI")ExIR Shiny app
Run ExIR, inspect result classes and explore candidate prioritization.
influential::runShinyApp("ExIR")
Quick start
Install it. Build a network. Ask what matters.
Start with the stable CRAN release, then move directly to the workflow that fits your question.
install.packages("influential")
library(influential)
# Development version
# install.packages("remotes")
remotes::install_github(
"asalavaty/influential",
build_vignettes = TRUE
)library(influential)
library(igraph)
g <- graph_from_data_frame(coexpression.data)
ivi_scores <- ivi(
graph = g,
directed = FALSE,
scale = "range"
)
head(ivi_scores)fit <- exir(
Diff_data = my_Diff_data,
Diff_value = c(1, 3),
Regr_value = 5,
Sig_value = c(2, 4),
Exptl_data = expression_matrix,
Exptl_data_type = "bulk",
condition = condition,
feature_filter = TRUE,
verbose = FALSE
)
names(fit)
# "Driver table"
# "DE-mediator table"
# "Biomarker table"
# "Graph"# Interactive IVI interface
influential::runShinyApp("IVI")
# Interactive ExIR interface
influential::runShinyApp("ExIR")Where it fits
From network science to experimental biology.
The same toolkit can be used to interrogate abstract networks or prioritize biologically meaningful candidates from high-dimensional experimental data.
Network influence analysis
Identify nodes that integrate local connectivity, neighborhood structure and global network position.
IVIDriver discovery
Prioritize features predicted to exert strong impact on the progression of a biological process or disease.
ExIR driversBiomarker prioritization
Rank features predicted to show high sensitivity to the conditions and severity represented in the experimental data.
ExIR biomarkersSingle-cell feature prioritization
Use raw-count single-cell inputs with pseudo-sampling, including sparse matrices and Seurat objects.
scRNA-seqInfluence benchmarking
Use simulation-based SIRIR ranks to assess how strongly individual nodes affect topology and spreading behavior.
SIRIRIn silico perturbation
Compare knockout and up-regulation effects on information flow before committing candidates to validation experiments.
comp_manipulateAssociated methods papers
Read the methodological foundations.
The package connects published network-influence methodology with newer experimental-data-driven feature prioritization.
ExIR enables prioritizing driver and biomarker genes from omics data in a reference free manner.
Salavaty A, Douek AM, Kaslin J, Ramialison M, Currie PD.
Read publication ↗Integrated Value of Influence: An Integrative Method for the Identification of the Most Influential Nodes within Networks.
Salavaty A, Ramialison M, Currie PD.
Read publication ↗Open & documented
Everything you need to install, learn and reproduce.
Use CRAN for the stable release, GitHub for development and issues, the vignette for step-by-step workflows, and the web portal for interactive analyses.
Package
Stable release, source code and issue tracking.
Documentation
Overview, detailed workflows and function references.
Interactive tools
Browser-based and locally launchable analysis interfaces.
What is the difference between influential and IVI?
influential is the R package. IVI is one of its featured network-influence methods. The package also contains ExIR, Hubness score, Spreading score, SIRIR, perturbation tools, network reconstruction, centrality functions and visualization utilities.
Can I use ExIR without starting from a pre-built network?
Yes. ExIR is designed around experimental data and can reconstruct feature relationships internally as part of its integrative prioritization workflow.
What kinds of experimental data can ExIR use?
The vignette describes transcriptomics, proteomics, bulk RNA-seq and single-cell RNA-seq applications. Inputs can include data frames, tibbles, matrices, sparse matrices and Seurat objects.
Does ExIR return only one ranked list?
No. Depending on the supplied evidence and arguments, ExIR can return a graph plus tables for drivers, biomarkers, DE-mediators and nonDE-mediators.
Can IVI handle directed or weighted networks?
Yes. IVI supports directed networks through the relevant direction/mode settings and can use edge weights where appropriate.
What is the difference between Hubness score and Spreading score?
Hubness score emphasizes local topological power using degree centrality and local H-index. Spreading score integrates measures associated with information-diffusion potential, including ClusterRank, neighborhood connectivity, betweenness and collective influence.
When should I use SIRIR?
Use SIRIR when you want a simulation-based view of influence. It combines SIR dynamics with leave-one-out perturbation to rank nodes according to their effect on network topology and information spread.
Can I use the package without coding?
The Influential Software Package web portal provides browser-based IVI and ExIR interfaces. The same apps can also be launched locally from R with runShinyApp("IVI") and runShinyApp("ExIR").
How do I cite the package?
Use the publication associated with the method you use. ExIR is described in the 2026 iScience article and IVI/network influence analysis is described in the 2020 Patterns article. citation("influential") provides package citation information from R.
Find the nodes and features that move the system.
Install influential from CRAN, explore the full vignette, or open the interactive IVI and ExIR interfaces.