enrichit: C++ Implementations of Functional Enrichment Analysis

enrichit is part of the clusterProfiler family, serving as the underlying algorithm implementation layer. It focuses on fast core computation, standardized result objects, and reusable data-preparation layers for downstream visualization packages such as enrichplot.

The package now covers not only classical enrichment workflows such as ORA and GSEA, but also weighted enrichment, network propagation-based enrichment, multi-omics early/late fusion, multi-layer topology fusion, and explanation-ready data extraction.

Installation

You can install the development version of enrichit from GitHub using devtools:

# install.packages("devtools")
devtools::install_github("YuLab-SMU/enrichit")

Scope

enrichit is designed around four layers:

• Core enrichment engines: ORA, GSEA, weighted ORA/GSEA, and GSON-aware variants.
• Network-aware enrichment: single-layer nsea() and multi-layer mnsea() workflows based on Random Walk with Restart.
• Multi-omics integration: early fusion at the feature level and late fusion at the pathway level.
• Explanation-ready outputs: contribution tables and topology-aware extraction helpers prepared for visualization in enrichplot.

Feature Map

• High performance core: key algorithms are implemented in C++ via Rcpp, with sparse network propagation powered by RcppEigen.
• ORA: standard hypergeometric ORA with optional weighted ORA through Wallenius' noncentral hypergeometric distribution.
• GSEA: multilevel, permutation, and adaptive strategies for ranked enrichment analysis.
• GSON support: native ora_gson() and gsea_gson() interfaces for structured gene set collections.
• NSEA: nsea() and nsea_gson() for network-ranked enrichment on a single graph, including mode = "signed" for bidirectional propagation.
• Multi-layer topology fusion: mnsea() and mnsea_gson() for multiplex or heterogeneous network propagation across multiple layers.
• Multi-omics early fusion: aggregate_omics(), harmonize_ids(), and select_features_for_ora() for feature-level integration before enrichment.
• Multi-omics late fusion: aggregate_enrichment() for pathway-level aggregation of multiple enrichment results.
• Contribution tracing: get_omics_contribution(), classify_omics_pattern(), and get_mnsea_contribution() for explanation-oriented summaries.
• Topology-aware extraction: extract_mnsea_subnetwork() for pathway-specific node/edge tables that can be passed to downstream visualization packages.
• Bayesian compression: bayes_enrich() and bayes_summary() for posterior-based term prioritization.

Main APIs

Classical enrichment

• ora(), ora_gson()
• gsea(), gsea_gson()
• gseaScores()

Weighted enrichment

• ora(..., weight = )
• ora_gson(..., weight = )
• gsea(..., weight = )
• gsea_gson(..., weight = )

Network-aware enrichment

• prepare_network()
• nsea(), nsea_gson()
• prepare_multilayer_network()
• propagate_multilayer()
• collapse_multilayer_scores()
• mnsea(), mnsea_gson()

Multi-omics integration

• aggregate_omics()
• harmonize_ids()
• select_features_for_ora()
• aggregate_enrichment()

Explanation helpers

• get_omics_contribution()
• classify_omics_pattern()
• get_mnsea_contribution()
• extract_mnsea_subnetwork()

Result Objects

The package provides the standard enrichment result object model used by the clusterProfiler family:

• enrichResult for ORA-like workflows
• gseaResult for ranked enrichment workflows
• nseaResult for single-network propagation plus enrichment
• mnseaResult for multi-layer propagation, collapsed scores, and cached explanation tables

These objects are intended to support a clean separation of concerns across the clusterProfiler family:

• enrichit handles core computation, algorithm implementation, and explanation-ready data preparation
• clusterProfiler provides high-level biological interpretation workflows and general enrichment analysis interfaces
• enrichplot handles visualization
• gson provides a structured gene set resource layer for managing and exchanging gene set collections across the family
• knowledge-base-oriented downstream packages such as DOSE, ReactomePA, meshes, and MicrobiomeProfiler provide domain-specific annotation and interpretation layers

Design Notes

• Computation first: this package prioritizes fast and robust numerical routines over plot helpers.
• Decoupled architecture: integration layers, propagation layers, and enrichment layers are exposed separately where useful.
• Stable downstream interface: explanation helpers return standard tables so that plotting logic can evolve independently in downstream packages.