Tissue-clocks

⬅️ read the preprint here
⬅️ cite the code

Histological aging signatures for monitoring tissue-specific aging and disease.

We develop histological aging clocks from H&E-stained tissue sections to predict biological age across multiple human tissues. These clocks capture tissue-specific aging signatures that associate with pathological outcomes, telomere length, and morbidity — providing a morphological window into the aging process. Transferring these clocks to blood gene expression profiles enables non-invasive prediction of tissue-specific aging and disease risk, highlighting the potential of histological features as biomarkers for aging and pathology. Licensed under PolyForm Noncommercial 1.0.0. Source code: Zenodo 10.5281/zenodo.20671352. Fine-tuning data: Zenodo 10.5281/zenodo.13330659.

---

🚀 Setup

This project uses uv for Python package and environment management, and just as a command runner.

To install dependencies:

uv sync

WSI-related scripts also require RendeiroLab/wsi (install separately).

To see available commands:

just

🔬 Running the analysis

Run the full original analysis pipeline:

just analysis

Run specific analysis stages:

just visualize   # Cohort and dataset inspection
just train       # Model fine-tuning
just features    # Histological feature extraction
just clocks      # Histological aging clocks
just rna         # Gene expression analysis
just dname       # DNA methylation analysis
just blood       # Blood prediction and associations
just validation  # Validation steps

Run revision work:

just revision    # All revision steps

Or run everything:

just analysis_full

---

📁 Code organization

The src directory contains all code used in the project. Not all steps are fully reproducible due to constraints in obtaining the data from dbGAP. Some steps while being reproducible are not feasible to be run in serial as they were run on multiple machines and configuring these is very environment dependent.

Below is an overview of all code used in the project.

📥 Getting data

• download_slides.sh
• download_gtex_dbGAP_restricted.sh
• download_gtex_dname.sh

⚠️ Not fully reproducible due to dbGAP constraints.

📊 Cohort and dataset inspection and QC

• visualize_sample_attributes.py
• visualize_morbidities.py
• visualize_pathology_data.py
• visualize_telomere_data.py

🧠 Model fine-tuning

• train_fastai_pre.py
• train_fastai.py
• train_fastai_post.py
• train_fastai_metrics.py

🔍 Histological feature extraction

• feature_extractor.py
• feature_extractor.job.sh
• feature_extractor.submit.py
• explore_histological_feature_space.py
• variance_explained_factors.py

⏰ Histological aging clocks (tissue-clocks)

• predict_histological_age.py
• visualize_age_gaps.py
• interpret_age_gaps_pathology.tissue.py
• interpret_age_gaps_telomeres.tissue.py
• interpret_age_gaps_across_tissues.py

💬 Interpretation through text

• interpret_age_gaps_text.py

🧬 Gene expression

• explore_rna_feature_space.py
• interpret_age_gene_expression.py

🧪 DNA methylation data

• explore_dname_feature_space.py
• predict_dname_age.py
• compare_dname_age.py

🔗 Associations

• associate_age_gaps_factors.py

🩸 Prediction from blood

• predict_gaps_from_blood_expression.py
• interpret_age_gaps_pathology.blood.py
• interpret_age_gaps_telomeres.blood.py

✅ Validation

• prepare_archs4_data.py
• validate_blood_predictors.py

🛠️ Cross project, supporting

• ops.py
• utils.py

📝 Revision work

• Cohorts:
  • cohort_description_table.py
• Vision models:
  • make_clocks_imagenet.py
• Foundation models and new cohorts:
  • process_lz.py
  • process_lz_aggregate.py
  • new_cohorts_clocks.py
  • cross_apply_clocks.py
  • correlate_histology_dname.py
  • Outcomes:
    • interpret_age_gaps_telomeres.tissue.imagenet.py
    • interpret_age_gaps_morbidity.tissue.imagenet.py
    • interpret_age_gaps_telomeres.tissue.py
    • interpret_age_gaps_pathology.tissue.py
    • interpret_age_gaps_morbidity.tissue.py
• GNNs:
  • gnn.prepare_graphs.py
  • gnn.train.py
  • gnn.train.per_organ.py
• DNAme comparison:
  • compare_dname-histology_telomeres.py
  • compare_dname-histology_pathology.py
  • compare_dname-histology_morbidity.py
  • compare_dname-histology_summary.py
• Specialized analysis:
  • compare_plip_conch.py
  • ecm_analysis.py
• Validation metrics:
  • icc_reproducibility.py - Within-tissue reproducibility via ICC (tile sampling)
  • icc_sample_tiles.py - Sample tiles from GTEx zarrs (run on remote)
  • icc_apply_clocks.py - Apply clocks to sampled tiles and compute ICC
  • bland_altman.py - Bland-Altman plots + calibration (slope/intercept/R²) with combined grid visualization
• Jupyter notebook:
  • make_notebook.py
  • tissue_clock_walkthrough.ipynb

📋 Supporting material

• supplementary_tables.py

---

📄 License

This repository is released under the PolyForm Noncommercial License 1.0.0. In plain terms:

• You may use, modify, and share this software for noncommercial purposes, including academic and nonprofit research.
• Commercial use is not permitted under this license. That includes use by for-profit organizations and use for commercial research, even if the work is described as research-only.
• If you need to use this software commercially, a separate license can be arranged through CeMM's intellectual property office: ip@cemm.at.

The full legal terms are in LICENSE.

---

📜 Citation

If you use this work, please cite the publication and this software archive:

Histological aging signatures for monitoring tissue-specific aging and disease Ernesto Abila, Iva Buljan, Yimin Zheng, Tamas Veres, Zhilong Weng, Maja Nackenhorst, Wolfgang Hulla, Yuri Tolkach, Adelheid Wöhrer, André F. Rendeiro bioRxiv 2024.11.14.618081. DOI: 10.1101/2024.11.14.618081 Software DOI: 10.5281/zenodo.20671352 TODO: Once the Nature Medicine DOI is available, update this block, the BibTeX entry below, and preferred-citation in CITATION.cff.

Full author list and metadata: CITATION.cff.

@article {Abila2024.11.14.618081,
	author = {Abila, Ernesto and Buljan, Iva and Zheng, Yimin and Veres, Tamas and Weng, Zhilong and Nackenhorst, Maja and Hulla, Wolfgang and Tolkach, Yuri and W{\"o}hrer, Adelheid and Rendeiro, Andr{\'e} F.},
	title = {Histological aging signatures for monitoring tissue-specific aging and disease},
	elocation-id = {2024.11.14.618081},
	year = {2026},
	doi = {10.1101/2024.11.14.618081},
	publisher = {Cold Spring Harbor Laboratory},
	URL = {https://www.biorxiv.org/content/early/2026/01/15/2024.11.14.618081},
	eprint = {https://www.biorxiv.org/content/early/2026/01/15/2024.11.14.618081.full.pdf},
	journal = {bioRxiv}
}

---

📬 Contact

For questions or collaborations, visit rendeiro.group or open an issue in this repository.