Microglial-Transcriptomics-AD

Single-cell RNA-seq analysis of microglial transcriptomics in Alzheimer's disease

Presentation

🔗 Google Slides:
Single-Cell RNA-Seq Analysis of Microglia in Alzheimer’s Disease

Overview

This repository contains a comprehensive single-cell RNA-sequencing (scRNA-seq) analysis validating key findings from Gatenby et al. (2025) published in Nature Medicine. The analysis demonstrates that amyloid-β immunization shifts microglial phenotype from metabolic stress to a neuroprotective state in Alzheimer's disease.

Key Findings

• nAD (non-immunized AD) microglia exhibit metabolic stress signature with upregulated mitochondrial genes (MT-ND1-6, MT-CYB, MT-CO2, MT-ATP6)
• iAD (immunized AD) microglia show restored protective phenotype with upregulated APOE, TREM2, CST3, and synaptic plasticity genes
• GO enrichment analysis confirms metabolic dysfunction → neuroprotection recovery pathway
• Results independently validate published findings using GSE263034 dataset

---

Repository Structure

Repository Structure

Microglial-Transcriptomics-AD/
│── data/                       # Raw & processed data (ignored in .gitignore)
│   ├── raw/                    # downloaded GSE files (optional)
│   └── processed/              # Seurat objects, RDS files (optional)
│
│── scripts/                    # All R scripts for analysis
│   ├── 01_qc_filtering.R
│   ├── 02_doublet_detection.R
│   ├── 03_sctransform_normalization.R
│   ├── 04_harmony_integration.R
│   ├── 05_clustering_umap.R
│   ├── 06_deg_analysis_mast.R
│   └── 07_go_enrichment.R
│
│── results/                    # Plots (UMAPs, volcano plots, GO plots)
│   ├── qc_plots/
│   ├── umap_plots/
│   ├── volcano_plots/
│   ├── go_plots/
│   └── tables/
│
│── README.md                   # Project documentation
│── LICENSE                     # MIT license
│── .gitignore                  # Prevents uploading large data files

---

Data

Source

• Dataset: GSE263034 (GEO Omnibus)
• Technology: Single-cell RNA-sequencing (10x Genomics)
• Samples: 25 samples
  • NND (Non-demented controls): 6 donors
  • nAD (Non-immunized AD): 6 donors
  • iAD (Immunized AD): 13 donors
• Cells analyzed: 106,298 singlets (post-QC, post-doublet removal)

Download

wget https://www.ncbi.nlm.nih.gov/geo/download/?acc=GSE263034&format=file

Methods & Workflow

1. Quality Control (QC) Filtering

• Filtered cells by:
  • nFeature_RNA: median ± 2 MAD per sample
  • nCount_RNA: median ± 3 MAD per sample
  • percent.mt > 20%
• Result: 106,298 high-quality cells retained (from 112,196)
• Cells removed: ~5,898 doublets + low-quality cells

2. Doublet Detection

• Used scDblfinder (per-sample detection, high-confidence mode)
• Removed predicted doublets while retaining singlets
• Result: 94.7% singlets retained (5,898 doublets removed)

3. Normalization

• Applied SCTransform normalization
• Regressed out mitochondrial percentage (percent.mt)
• Stabilized variance across genes
• Selected highly variable genes (HVGs) for downstream analysis

4. Integration & Batch Correction

• Method: Harmony (batch correction)
• Combined all 25 samples from different donors
• PCA: Computed 50 PCs, selected 30 based on elbow plot
• Corrected for donor-specific batch effects
• Result: Samples now cluster by biology, not batch

5. Clustering & Cell Type Identification

• Algorithm: Seurat-based clustering (Louvain method)
• Dimensionality: UMAP visualization
• Clusters identified: 27 distinct cell clusters
• Filtered to microglial populations (g23, g21, g20) for differential expression

6. Differential Expression Analysis

• Method: MAST (Model-based Analysis of Single-cell Transcriptomics)
• Statistical test: Two-sample comparison with covariate adjustment
• Comparisons performed:
  • nAD vs NND: Disease stress signature in non-immunized AD
  • iAD vs nAD: Immunization rescue effect
• Significance thresholds:
  • log2 Fold Change > 0.5 (or < -0.5)
  • Adjusted p-value < 0.05

7. Gene Ontology (GO) Enrichment Analysis

• Package: clusterProfiler
• Ontology: Biological Process (GO-BP)
• Background: All genes expressed in dataset
• Enrichment method: Hypergeometric test
• Cutoff: padj < 0.05
• Result: Identified metabolic stress and neuroprotective pathways

---

Data Analysis Workflow Summary

Raw Data (GSE263034)
↓
QC Filtering (nFeature, nCount, percent.mt)
↓
Doublet Detection (scDblfinder)
↓
Normalization (SCTransform)
↓
Integration (Harmony)
↓
Clustering (Louvain + UMAP)
↓
Cell Type Selection (Microglial subset)
↓
Differential Expression (MAST)
↓
GO Enrichment (clusterProfiler)
↓
Validation & Interpretation

Key Results

Volcano Plot: nAD vs NND

UP-regulated in nAD:

• MT-ND1, MT-ND2, MT-ND3, MT-ND4, MT-ND5, MT-ND6
• MT-CYB, MT-CO2, MT-ATP6
• Indicates metabolic stress and compromised oxidative phosphorylation

DOWN-regulated in nAD:

• TMSB4X (actin regulation)
• Indicates reduced cytoskeletal support

Volcano Plot: iAD vs nAD

UP-regulated in iAD:

• APOE, TREM2, CST3, S100A1 (neuroprotection)
• CKB, FAM107A (metabolic support)
• Indicates protective microglial phenotype

DOWN-regulated in iAD:

• HSPA1A, HSPA1B, HSP90AA1 (heat shock proteins)
• HSPH1, PTGDS
• Stress response pathways turned OFF

GO Enrichment: nAD

• Mitochondrial organization
• ATP metabolic processes
• Macroautophagy
• Interpretation: nAD microglia are highly stressed, working hard to manage damage

GO Enrichment: iAD

• Axonogenesis
• Neuron projection development
• Cognition
• Synaptic plasticity
• Interpretation: iAD microglia recovered neuroprotective function

---

Usage

Prerequisites

Install required packages install.packages(c("Seurat", "harmony", "clusterProfiler", "org.Hs.eg.db")) BiocManager::install(c("DESeq2", "MAST", "scDblFinder"))

---

Key Genes & Pathways

Metabolic Stress Signature (nAD)

Gene	log2FC	padj	Function
MT-ND1	3.2	1.2e-50	Complex I, OXPHOS
MT-ND2	2.8	3.4e-45	Complex I, OXPHOS
MT-CYB	2.5	2.1e-40	Complex III, OXPHOS
MT-ATP6	3.5	8.9e-65	ATP synthase

Neuroprotective Signature (iAD)

Gene	log2FC	padj	Function
APOE	1.8	1.2e-35	Amyloid clearance, lipid metabolism
TREM2	1.2	4.5e-28	Microglial activation, phagocytosis
CST3	2.1	3.7e-42	Neuroprotection, neuroinflammation suppression
S100A1	1.5	2.2e-30	Calcium signaling, synaptic plasticity

---

Validation

This analysis independently validates findings from:

Gatenby et al. (2025). "Microglial mechanisms drive amyloid-β clearance in immunized patients with Alzheimer's disease." Nature Medicine, 31, 1604–1616.
DOI: 10.1038/s41591-025-03574-1

Reproduced findings:

• Mitochondrial gene upregulation in nAD microglia
• APOE/TREM2/CST3 upregulation in immunized AD microglia
• Heat shock protein downregulation following immunization
• GO enrichment patterns confirm metabolic stress → neuroprotection recovery

---

Skills & Technologies Demonstrated

Bioinformatics Tools & Languages:

• R programming (Seurat, Harmony, clusterProfiler, MAST)
• Single-cell RNA-seq analysis & visualization
• Statistical analysis (differential expression, GO enrichment)
• HPC cluster computing (bash, job submission)
• Git & GitHub for version control

Computational Biology Expertise:

• Quality control & preprocessing (QC filtering, doublet detection, normalization)
• Batch effect correction (Harmony integration)
• Cell clustering & annotation (UMAP, Louvain clustering)
• Differential expression analysis (MAST statistical testing)
• Functional annotation (Gene Ontology enrichment)

Research Skills:

• Literature review & paper validation
• Independent analysis & reproducibility
• Scientific interpretation & communication
• Presentation design & data visualization

---