- An, L., Pichet Binette, A., Hristovska, I., Vilkaite, G., Xiao, Y., Smets, B., ... & Vogel J.W., (2026) A deep joint-learning proteomics model for diagnosis of six conditions associated with dementia, Nature Medicine.
The rapid growth of the dementia population underscores the urgent need for accessible, scalable biomarkers of neurodegenerative diseases. Such biomarkers are crucial for improving early diagnosis, confirming (co-)pathologies, and optimizing participant selection in clinical trials. This highlights the need for a one-shot, multi-disease biomarker panel, while emerging high-throughput plasma proteomics technologies offer promising avenues. In this work, we developed ProtAIDe-Dx to analyze complex, high-dimensional proteomics data from GNPC, the largest neurodegenerative disease proteomics dataset to date (N = 17,187). We hope this work establishes a robust benchmark for AI-driven proteomics approaches and supports precision medicine in neurodegenerative diseases.
Our project strucrture is as follows:
An_ProtAIDe-Dx/
├── checkpoints/ # Fitted models
│ ├── data_proc/ # Fitted data processing pipelines
│ └── ProtAIDeDx/ # Fitted ProtAIDe-Dx models (cross-validation and leave-one-site out)
├── data/ # Simulated data and metadata used by the pipeline
├── example/ # Minimal runnable examples and demo Jupyter notebook
├── replication/ # Scripts/configs to reproduce main results
└── src/ # Source code
- Our code uses Python, here is the setup procedure:
i. Install Miniconda or Anaconda with Python 3.x if you don't have conda installed;
ii.Create a conda environment usingconda env create -f replication/config/DeMONLab_ProtAIDe_env.yaml -n ProtAIDe.
- The example of code usage is detailed in
example/README.md
- Detailed step-by-step replication instructions are provided in
replication/README.md
- Release v1.0 (21/01/2025): Initial release of An_ProtAIDe-Dx project.
Please open a GitHub issue for any bugs or questions about the code. We will do our best to respond as soon as possible.