Skip to content

Killifish Transcriptome Aging Clock

Python adaptation of the KillifishAtlas aging clock pipeline for AAlab. Applies three transcriptomic aging clocks — BayesAge 2.0, Elastic Net (EN), and Principal Component Regression (PCR) — to Nothobranchius furzeri RNA-seq data and predicts transcriptomic age (tAge) in query samples.

What this repo does

  1. Trains three aging clocks on the published KillifishAtlas reference dataset (677 samples, 15 tissues, ages 47–163 days).
  2. Transfers the trained models to novel query datasets via ComBat-seq batch correction.
  3. Predicts transcriptomic age (tAge) for each query sample.
  4. Visualizes predictions as scatter-box plots and generates per-gene expression profiles across tissues and ages.
  5. Preprocessor of RNAseq for Killifish process other public dataset.

Repository layout

src/                        # Core pipeline modules
├── data_loader.py          # Atlas counts, metadata, query file loading
├── normalization.py        # Frequency and DESeq2 normalization
├── preprocessing.py        # Gene filtering, stratification, outlier detection
├── bayesage2.py            # BayesAge 2.0 clock
├── elastic_net.py          # Elastic Net clock
├── pcr.py                  # Principal Component Regression clock
├── calibration.py          # Apply clocks to query datasets
└── gene_mapping.py         # Map Ensembl IDs → Atlas gene names

run_query_clocks.py         # General-purpose CLI: apply clocks to any count matrix

plot_gene_expression.py     # Per-gene TPM expression across tissues + timecourse

data/                       # Gene mapping tables and reference GTF
raw_RNAseq_process/         # Total RNA-seq: download → QC → align → count matrix
unittests/                  # pytest test suite (8 test files)
Costa_et_al/                # Original KillifishAtlas analysis scripts
data/                       # Atlas TPM/count matrices, metadata, and gene mapping
query_data/                 # Query input files
outputs/                    # Generated prediction and figure outputs

Quick start

conda env create -f environment.yml
conda activate killifish-tx-clock

# 1. Normalize Atlas reference (once)
python src/normalize_reference.py

# 2. Run all three clocks on any count matrix
python run_query_clocks.py --counts query_data/toy.csv

# 3. Plot a gene's expression across tissues
python plot_gene_expression.py actb --log --timecourse

See Setup and Pipeline Overview for full details.