Single-cell proteomics of tau-bearing neurons in Alzheimer's disease (Foiani, Bourdenx, Kraller, et al. 2026)

This repository contains the analysis code for the single-cell and mini-bulk (pooled ~20 cells) proteomics study of laser-capture microdissected (LCM) neurons from human Alzheimer's disease frontal cortex. Neurons were stratified by phospho-tau (pTau) immunofluorescence status prior to microdissection and analysed by DIA-PASEF mass spectrometry on a timsTOF Ultra.

Data availability

Raw data and results are embargoed until publication and are therefore excluded from this repository via .gitignore. Processed AnnData objects and raw intensity matrices will be deposited in a public repository upon publication.

Repository structure

├── notebooks/                   # Analysis notebooks (run in order)
│   ├── 0-Preprocessing.ipynb    # AnnData assembly, QC, filtering
│   ├── 1-Figure1.ipynb          # Protein detection, dynamic range, neuronal markers
│   ├── 1-2-TauProfile.ipynb     # Tau peptide detection profile across 2N4R sequence
│   ├── 2-Figure2_minipools.ipynb# Mini-bulk: pseudotime, correlation, GSEA, heatmaps
│   ├── 3-Figure2_singlecells.ipynb # Single-cell: tau correlation, clustering, cell death
│   ├── 4-Figure3_proteostasis.ipynb # Proteasome/lysosome scores, K48-ubiquitin, V-ATPase
│   └── 5-Figure4_breakpoint.ipynb   # Piecewise regression / breakpoint analysis
├── src/                         # Shared Python modules and R scripts
│   ├── plot_all_datasets.py     # Cross-modality protein plotting with ANOVA + Tukey HSD
│   ├── pathway_score.py         # Z-score-based pathway scoring
│   ├── breakpoint_analysis.py   # Robust z-score and piecewise regression per protein
│   └── EWCE.R                   # Cell-type enrichment analysis (requires R, see below)
├── data/                        # Raw and processed data (not tracked)
│   ├── raw_data/                # Perseus exports, metadata, demographics, ubiquitin
│   └── processed_data/          # .h5ad files, EWCE results, breakpoint pickle
├── results/                     # Supplementary tables and gene lists (not tracked)
└── figures/                     # Output figures organised by panel

Dependencies

Tested with Python 3.12.

Python

Package	Version	Purpose
anndata	0.11.4	Data structure (AnnData / .h5ad)
scanpy	1.11.1	Single-cell analysis toolkit
pandas	2.2.3	Data manipulation
numpy	2.2.5	Numerical computing
scipy	1.15.2	Statistics (correlation, FDR, clustering, interpolation)
statsmodels	0.14.4	ANOVA, post-hoc tests
matplotlib	3.10.1	Plotting
seaborn	0.13.2	Statistical visualisation
PyComplexHeatmap	1.8.2	Annotated clustered heatmaps
adjustText	1.3.0	Non-overlapping text labels
palantir	1.4.1	Pseudotime inference (diffusion maps)
gseapy	1.1.11	Gene set enrichment analysis (pre-ranked GSEA, Enrichr)
scikit-learn	1.5.2	Dimensionality reduction
piecewise-regression	1.5.0	Breakpoint / segmented regression
joblib	1.4.2	Parallel computation
tqdm	4.67.1	Progress bars
openpyxl	3.1.5	Excel file reading

R

src/EWCE.R performs cell-type enrichment analysis and requires R with the following packages: EWCE, MAGMA.Celltyping, reticulate, SingleCellExperiment, scKirby.

Setup

Requires Python >= 3.11. Use uv for environment setup:

uv venv
source .venv/bin/activate
uv pip install -e ".[notebooks]"

Running the analysis

Start a Jupyter session from the project root:

jupyter lab

Then open notebooks in notebooks/ and run them sequentially. The preprocessing notebook (0-Preprocessing.ipynb) generates the .h5ad files consumed by all downstream notebooks.

Author

Mathieu Bourdenx — UK Dementia Research Institute at UCL