Featured project
Predicting Biological Age from Single-cell Transcriptomics
CSIC-IBMB | Sep 2025 - Dec 2025
Predict biological age from scRNA-seq data in both regression (age in months) and classification (young vs adult) setups.
Datasets
- Tabula Muris Senis: 23 organs, ~529k cells, ages 1/3/18/21/24/30 months.
- Whelan Esophagus: ~90k epithelial basal cells, young (<3m) vs aged (>19m).
Methods
- Regression: Elastic Net, Random Forest Regressor, XGBoost.
- Classification: Logistic Regression, Random Forest, SVM.
- Preprocessing: highly variable genes and correlation with age.
Results
- Best regression R2 reached 0.309, showing moderate age signal capture.
- Classification performance was clearly stronger for broad age bins.
- Feature importance highlighted genes linked to inflammaging and mitochondrial decline.
Future work
- Increase sample size and age balance.
- Mitigate batch effects across tissues.
- Explore pseudotime modeling and multi-omics integration.
More projects
Selected work from the portfolio database
No additional projects loaded yet. The featured CSIC project above is currently the main case study.