AMP-PD Single-Cell Analysis Pipeline

This repository contains scripts and notebooks used for the analysis of single-nucleus RNA-sequencing (snRNA-seq) data in the AMP-PD cohort. The project includes data preprocessing, clustering, cell type composition modeling, GWAS scoring, differential expression, TF inference, Hotspot module detection, and more — supporting our accompanying manuscript.

---

Table of Contents

• Project Overview
• Preprocessing & Clustering
• GWAS Scoring with scDRS
• Cell Type Compositional Analysis
• Differential Expression with dreamlet
• Variance Partitioning
• Hotspot Module Detection
• TF Activity Inference
• Cell-Cell Interaction Inference
• Installation & Requirements

---

Project Overview

This pipeline supports the full spectrum of single-cell and pseudobulk-level analyses in the AMP-PD dataset. Modules include:

• Multistage data preprocessing, QC, and integration (Scanpy + Harmony)
• Cell-type proportion modeling across clinical groups (crumblr + dream)
• GWAS signal scoring with scDRS
• Differential expression via dreamlet with hierarchical covariates
• Hotspot-based gene module discovery
• TF activity inference and regulatory network visualization (decoupler)
• Cell-cell interaction prediction (LIANA)

---

Preprocessing & Clustering

Script: prepare_and_integrate_AMPPD.py
Performs HVG filtering, PCA, Harmony batch correction, and Leiden clustering using Scanpy and Pegasus.

Run

python src/prepare_and_integrate_AMPPD.py

---

GWAS Scoring with scDRS

Script: score_gwas_scdrs.py
Scores single cells using scDRS for prioritized Parkinson’s disease GWAS gene sets.

Run

python src/score_gwas_scdrs.py

---

Cell Type Compositional Analysis

Script: cell_type_composition.R
Performs compositional modeling across subclasses using crumblr and dream, followed by meta-analysis with metafor and visualization via ggtree.

Output

• Meta-analysis of compositional shifts
• Coefficient plots annotated on hierarchical cell tree

---

Differential Expression with dreamlet

Script: dreamlet_differential_expression_PD.R
Runs subclass-level pseudobulk DE using dreamlet. Includes models with and without ethnicity/participant covariates.

Output

• DE results in .csv and .RData

---

Variance Partitioning

Script: variance_partition_braakLB_dreamlet.R
Uses dreamlet::fitVarPart() to quantify gene expression variance explained by covariates including Braak LB stage.

Run

Rscript scripts/variance_partition_braakLB_dreamlet.R

---

Hotspot Module Detection

Script: hotspot_example.py
Runs the Hotspot algorithm to detect local autocorrelated gene modules, focusing on myeloid cells.

Run

python src/hotspot_example.py

---

TF Activity Inference

Script: tf_inference.py
Infers transcription factor activity per cell type using decoupler and CollecTRI network. Includes specificity scoring, normalized activity, and heatmap visualization.

Output

• Normalized TF activity scores
• TF specificity scores
• Ranked TFs per subclass with heatmap

---

Cell-Cell Interaction Inference

Script: run_cci_liana_example.py
Uses the LIANA Python package to infer cell-cell interactions from integrated single-cell data.

Run

python src/run_cci_liana_example.py

---

Installation & Requirements

Python (via pip)

The following packages are required and can be installed via pip:

pip install scanpy pegasuspy pegasusio anndata scdrs matplotlib seaborn pandas numpy \
    decoupler liana igraph scikit-learn harmony-pytorch

Note:

• pegasuspy and pegasusio are required for HDF5/AnnData processing
• harmony-pytorch is used for batch correction
• decoupler is used for TF activity inference
• liana is used for CCI prediction

---

R (via Bioconductor + CRAN)

Use the following commands to install the required R packages:

# Bioconductor
BiocManager::install(c(
  "zellkonverter", "SingleCellExperiment", "dreamlet", "variancePartition",
  "crumblr", "ggtree", "qvalue", "GSEABase", "BiocParallel"
))

# CRAN
install.packages(c(
  "ggplot2", "tidyverse", "aplot", "broom", "cowplot", "metafor", "reticulate"
))

---