BrainEnrich is an R package designed to facilitate the integration of brain imaging data with transcriptomic profiles. It enables researchers to explore the molecular underpinnings of brain phenotypes by performing enrichment analysis of predefined gene sets. Whether working at the group or individual level, the package offers a flexible and powerful tool for examining associations between brain imaging phenotypes (e.g., cortical thickness) and gene expression, using a variety of statistical models, null models, and aggregation methods.

Overview of the BrainEnrich package and its analysis workflow.

🚀Features

• A. Group-Level Enrichment Analysis: A group-level IDP (e.g., effect size maps) is correlated with AHBA transcriptional profiles, yielding a spatial coupling vector that is aggregated within the predefined gene sets. The resultant "GS score" vector is tested against the null GS scores generated by the null models.Tutorial: Group-level Enrichment Analysis
• B. Individual-Level Enrichment Analysis: A similar GS scoring approach is applied to individual IDPs, generating a spatial coupling matrix that is aggregated within the predefined gene sets. The resultant participant-specific GS scores are used for downstream analyses, and the results are tested against the null GS scores generated by the null models. Tutorial: Individual-level Enrichment Analysis
• C. Predefined Gene Sets: Predefined gene sets from various sources and user-defined sets aggregate gene-specific associations into GS scores for either group-level or individual-level analyses. The package includes predefined gene sets like: Gene Ontology, DisGeNet, KEGG, WikiPathways, Reactome, MeSH, SynGO, and Cell Type gene sets. Tutorial: How the predefined gene sets are prepared
• D. Null Models: Includes both self-contained (spinning brain regions) and competitive (resampling genes) null models to assess the significance of GS score or to generate null GS scores at individual level for downstream analysis.
• E. Simulation Studies: Type I error and power simulations to assess the reliability of the analysis methods. Tutorial: Simulation
• F. Core Genes Identification: A leave-one-out (LOO) procedure identifies genes that substantially influence GS scores (in group-level analysis) or the resultant test statistic (in individual-level analysis), highlighting the primary contributors to enriched terms.
• 🔧Multiple Association Methods: Pearson and Spearman correlations, Partial Least Squares (PLS) regression, and user-defined methods for exploring gene-imaging associations.[Tutorial:Customize Association Methods (in prep)]
• 🔧Aggregation Methods: Multiple options for aggregating gene set scores, including mean, median, and Kolmogorov-Smirnov (KS)-based statistics.[Tutorial:Customize Aggregation Methods (in prep)]
• �︎ HCP (Slurm): Provides a detailed tutorial on running brainscore and simulation studies on high-performance computing systems using Slurm, enabling efficient large-scale analyses.Tutorial:Computation on HPC with slurm

💾Installation

# Install remotes if you haven't already
if (!requireNamespace("remotes", quietly = TRUE)) {
      install.packages("remotes")}
if (!requireNamespace("BiocManager", quietly = TRUE)) {
    install.packages("BiocManager")
}
BiocManager::install("DOSE")
# Install brainEnrich from GitHub
remotes::install_github("zh1peng/BrainEnrich")

Note that the pacakge will be available on CRAN soon.

🔬Usage

Tutorials are available at github-pages

👨‍💻Author

• Zhipeng Cao @ Xuhui Mental Health Center, Shanghai - Initial work - zh1peng

📜License

This project is licensed under the GNU Affero General Public (AGP) License - see the LICENSE.md file for details. We welcome contributions to BrainEnrich! Feel free to fork the repository, and create pull requests.