Brain Multi-Omics Analysis Pipeline

A reproducible R-based pipeline for end-to-end analysis of DIA proteomics and bulk RNA-seq transcriptomics data from mouse or human brain tissue. Originally developed for Alzheimer's disease mouse models but generalized for any case-control multi-omics study with multiple genotypes, sexes, and timepoints.

---

📋 What This Pipeline Does

Notebook 01 — DIA Proteomics Pipeline

Step	Analysis
QC	Protein abundance matrix loading, boxplot QC, summarization to protein level
PCA	Probabilistic PCA (handles missing values) with metadata overlay
Differential Abundance	ANOVA + Tukey HSD post-hoc, stratified by sex and age
Visualization	Volcano plots, DEP barplots, UpSet overlap plots, expression boxplots
Functional Enrichment	KEGG pathway ORA via clusterProfiler

Notebook 02 — Proteomics Human Translation Analysis

Step	Analysis
Linear Modeling	abundance ~ Genotype + Age + Sex per protein with covariate control
Module Correlation (logFC)	Pearson correlation of ANOVA fold-changes with human AD co-expression modules
Module Correlation (LM)	Pearson correlation of LM estimates with human AD co-expression modules
BioDomain ORA	GO over-representation analysis annotated with AMP-AD biological domains
BioDomain GSEA	Full ranked GSEA annotated with AMP-AD biological domains
Sub-BioDomain	Portrait and landscape plots across AD biological sub-domains

Notebook 03 — RNA-seq Transcriptomics Pipeline

Step	Analysis
Data Inspection	Metadata loading, count matrix QC, transgene merging (optional)
Isoform Expression	Visualization of specific gene isoforms e.g. human MAPT (optional)
Sex Validation	Xist / Ddx3y expression to confirm sex assignments
PCA	DESeq2 VST-based PCA + PCAtools biplot and eigencorplot
Differential Expression	DESeq2 pairwise comparisons, stratified by sex and age
Visualization	Volcano plots, DEG barplots
Functional Enrichment	KEGG pathway ORA via clusterProfiler

Notebook 04 — RNA-seq Human Translation Analysis

Step	Analysis
Module Correlation (logFC)	Pearson correlation of DESeq2 logFC with sex-stratified AMP-AD modules
Linear Modeling	Per-gene LM on TPM data with genotype, sex, and age covariates
Module Correlation (LM)	Pearson correlation of LM estimates with AMP-AD modules
AD Subtype Correlation	Correlation with Nikhil et al. and Neff et al. AD subtype signatures
BioDomain Correlation (LM)	LM estimates vs AMP-AD BioDomain reference fold-changes
BioDomain ORA	GO over-representation annotated with AMP-AD biological domains
BioDomain GSEA	Full ranked GSEA annotated with AMP-AD biological domains

---

🗂️ Repository Structure

brain-multiomics-pipeline/
│
├── README.md                                        # This file
├── LICENSE                                          # MIT license
├── .gitignore                                       # Excludes data, results, and large files
│
├── notebooks/
│   ├── 01_DIA_Proteomics_Pipeline.Rmd               # Proteomics QC, PCA, ANOVA, KEGG
│   ├── 02_Human_Translation_Analysis.Rmd            # Proteomics LM, modules, BioDomain
│   ├── 03_RNAseq_Pipeline.Rmd                       # RNA-seq QC, PCA, DESeq2, KEGG
│   └── 04_RNAseq_Human_Translation.Rmd              # RNA-seq LM, modules, subtypes, BioDomain
│
├── R/
│   ├── Helper_Functions_Prot.R                      # Proteomics utility functions
│   └── Helper_Functions.R                           # Transcriptomics utility functions
│
├── data/
│   └── example/
│       ├── example_abundance.csv                    # Example protein abundance matrix
│       └── example_traits.csv                       # Example sample metadata
│
├── figures/                                         # Example output figures
│
├── results/                                         # gitignored — outputs saved here locally
│
└── docs/
    └── input_format.md                              # Detailed input file format guide

Run order: Notebooks must be run in sequence within each omics type. Notebook 02 requires outputs from 01. Notebook 04 requires outputs from 03. The proteomics and transcriptomics pipelines are independent of each other.

---

⚙️ Requirements

R Version

R ≥ 4.2.0

CRAN Packages

install.packages(c(
  "tidyverse", "gt", "UpSetR", "data.table", "broom",
  "ggpubr", "ggh4x", "corrplot", "cowplot", "ggrepel",
  "matrixStats", "VennDiagram", "readxl", "readr"
))

Bioconductor Packages

if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install(c(
  # Core
  "AnnotationDbi",     # Gene ID mapping
  "clusterProfiler",   # Pathway enrichment (ORA and GSEA)
  "enrichplot",        # Enrichment visualization
  "ComplexHeatmap",    # Heatmap visualization

  # Proteomics specific
  "pcaMethods",        # Probabilistic PCA (handles missing values)

  # Transcriptomics specific
  "DESeq2",            # Differential expression
  "PCAtools",          # Advanced PCA and eigencorplot
  "DEGreport",         # DEG reporting utilities
  "ReactomePA",        # Reactome pathway analysis
  "limma",             # Linear models for omics

  # Organism annotation — install only what you need
  "org.Mm.eg.db",      # Mouse
  "org.Hs.eg.db",      # Human
  "org.Rn.eg.db"       # Rat
))

Species note: Only one organism package is needed per study. Set organism_db in the USER CONFIGURATION block of each notebook.

Optional — Human Translation (Notebooks 02 and 04)

External reference datasets required for human comparison analyses:

Reference	Source	Used in	Used for
AMP-AD co-expression modules	AMP-AD Knowledge Portal	02, 04	Module logFC and LM correlation
AMP-AD sex-stratified modules	AMP-AD Knowledge Portal	04	Sex-specific module correlation
BioDomain annotations	Agora / AMP-AD	02, 04	GO term → BioDomain mapping
Domain color table	AMP-AD	02, 04	BioDomain plot styling
Milind et al. AD subtypes	Milind et al. 2021	04	AD subtype correlation
Neff et al. AD subtypes	Neff et al. 2021	04	AD subtype correlation

---

🚀 Quick Start

Proteomics Pipeline (Notebooks 01 → 02)

Step 1: Prepare your input files (see docs/input_format.md):

• Normalized protein abundance matrix (CSV)
• Sample metadata (CSV)

Step 2: Configure and run notebook 01:

# Key settings in the USER CONFIGURATION block:
data_dir         <- "../data/"
abundance_file   <- "your_abundance_file.csv"
traits_file      <- "your_metadata_file.csv"
result_dir       <- "../results/"

control_genotype <- "Your_Control_Genotype"
case_genotypes   <- c("Case_1", "Case_2", "Case_3")
age_groups       <- c(4, 12)
organism_db      <- "org.Mm.eg.db"    # swap for human
kegg_organism    <- "mmu"             # "hsa" for human

rmarkdown::render("notebooks/01_DIA_Proteomics_Pipeline.Rmd")

Step 3: Configure and run notebook 02 (optional):

# Enable only the analyses you have reference data for
run_module_corr    <- TRUE
run_biodomain_ora  <- TRUE
run_biodomain_gsea <- TRUE
run_lm_analysis    <- TRUE

rmarkdown::render("notebooks/02_Human_Translation_Analysis.Rmd")

---

Transcriptomics Pipeline (Notebooks 03 → 04)

Step 1: Prepare your input files:

• RSEM gene count matrix (TSV)
• RSEM gene TPM matrix (TSV)
• Sample metadata (CSV)
• Optionally: isoform TPM matrix (TSV)

Step 2: Configure and run notebook 03:

# Key settings in the USER CONFIGURATION block:
data_dir           <- "../data/"
gene_counts_file   <- "rsem.merged.gene_counts.tsv"
gene_tpm_file      <- "rsem.merged.gene_tpm.tsv"
metadata_file      <- "metadata_validated.csv"
result_dir         <- "../results/"

control_genotype   <- "Your_Control_Genotype"
case_genotypes     <- c("Case_1", "Case_2", "Case_3")
age_groups         <- c(4, 12)
organism_db        <- "org.Mm.eg.db"
kegg_organism      <- "mmu"

# Transgene handling — set TRUE only if your model has human transgenes
has_transgenes     <- FALSE

rmarkdown::render("notebooks/03_RNAseq_Pipeline.Rmd")

Step 3: Configure and run notebook 04 (optional):

# Enable only the analyses you have reference data for
run_module_corr_fc <- TRUE
run_module_corr_lm <- TRUE
run_subtype_corr   <- TRUE
run_lm_analysis    <- TRUE
run_biodomain_ora  <- TRUE
run_biodomain_gsea <- TRUE

rmarkdown::render("notebooks/04_RNAseq_Human_Translation.Rmd")

---

📁 Input File Formats

Proteomics — Abundance Matrix

ProteinID,Sample1,Sample2,Sample3,...
MAPT|peptide1,12.3,11.8,12.1,...
APP|peptide2,10.5,10.2,10.8,...

• Rows = proteins; columns = samples
• Values should be pre-normalized (e.g., TAMPOR-corrected, log2-scale)
• ProteinID can be Protein or Protein|Peptide format

All Notebooks — Sample Metadata

X,Sex,Genotype,Age
Sample1,Female,Control,4
Sample2,Male,Control,4
Sample3,Female,CaseA,4

• X: Sample ID — must exactly match column names in abundance/count matrix
• Sex: Male / Female
• Genotype: group labels matching your config variables
• Age: numeric timepoint in months

Transcriptomics — Count Matrix

Standard RSEM TSV output with gene_id as first column and one column per sample. See docs/input_format.md for full format details and edge cases.

---

🧪 Test with Example Data

# For proteomics notebook 01, set in USER CONFIGURATION:
data_dir       <- "../data/example/"
abundance_file <- "example_abundance.csv"
traits_file    <- "example_traits.csv"

Example data contains 20 proteins and 12 samples across 2 genotypes, 2 sexes, and 2 ages. Designed to verify end-to-end pipeline execution — enrichment results are not biologically meaningful at this scale.

---

📊 Output Files

Output	Location	Generated by	Description
HTML Reports	notebooks/	All notebooks	Full rendered analysis with all plots
Processed Proteomics Data	data/Processed_DIA_Proteomics.RData	Notebook 01	Cleaned abundance matrix + traits
ANOVA Results	results/ANOVA_Results_DIA_Proteomics.Rdata	Notebook 01	All DEP results by comparison
Proteomics Human Translation	results/Human_Translation_Results.Rdata	Notebook 02	LM, module, BioDomain results
Processed RNA-seq Data	data/ProcessedData_RNAseq.RData	Notebook 03	Filtered count + TPM matrices
DESeq2 Results	results/DESeq2_Results_RNAseq.RData	Notebook 03	All DEG results by comparison
LM Results (RNA-seq)	results/LM_Results_RNAseq.RData	Notebook 04	LM estimates by gene and variant
ORA BioDomain (RNA-seq)	results/ORA_BioDomain_RNAseq.RData	Notebook 04	GO-ORA annotated with BioDomains
GSEA BioDomain (RNA-seq)	results/GSEA_BioDomain_RNAseq.RData	Notebook 04	GO-GSEA annotated with BioDomains
RNA-seq Human Translation	results/Human_Translation_RNAseq.RData	Notebook 04	All human correlation results

---

🔬 Methods Summary

Proteomics QC: Sample distributions assessed by boxplot after normalization. Proteins with pipe-delimited IDs summarized to protein level by averaging duplicate entries. Probabilistic PCA (pcaMethods::pca()) accommodates missing values common in DIA data.

Differential Abundance (ANOVA): One-way ANOVA per protein with Tukey HSD post-hoc correction, stratified by sex and age group. Only group pairs present in the data are tested — invalid comparisons are automatically skipped.

Transcriptomics QC: Optional transgene merging for models with human gene inserts (configurable via lookup table). Sex validated using Xist and Ddx3y expression. Low-count genes pre-filtered by minimum TPM threshold.

Differential Expression (DESeq2): Pairwise DESeq2 comparisons per sex and age stratum using raw integer counts. Genes annotated with symbols and Entrez IDs via AnnotationDbi.

Linear Modeling: Per-gene/protein linear models estimate the independent contribution of each genotype while controlling for covariates (Age, Sex). Model formula and coefficients are fully user-configurable. Stratified by age group where relevant.

Pathway Enrichment: KEGG over-representation analysis using clusterProfiler::compareCluster(). Gene set enrichment analysis (GSEA) using clusterProfiler::gseGO() on full ranked gene lists.

Human Translation: Mouse model fold-changes and LM estimates correlated (Pearson) with human AMP-AD co-expression module signatures (sex-stratified and combined). GO enrichment results mapped to the AMP-AD BioDomain framework across 19 AD biological domains and sub-domains. AD subtype correlations against Nikhil et al. (ROSMAP, Mayo, MSBB) and Neff et al. (MSBB) reference datasets.

---

📖 Citation

If you use this pipeline, please cite:

Pandey RS. Brain Multi-Omics Analysis Pipeline. GitHub: https://github.com/pandeyravi15/brain-multiomics-pipeline

And the underlying tools:

• DESeq2
• clusterProfiler
• EnhancedVolcano
• pcaMethods
• PCAtools
• AMP-AD BioDomain Framework

---

🤝 Contributing

Issues and pull requests are welcome. Please open an issue first to discuss major changes.

---

📄 License

MIT — see LICENSE for details.

---

📬 Contact

Ravi S Pandey
GitHub: @pandeyravi15