Spatial mRNA Decay

Infers spatially-varying mRNA degradation rates from eve stripe transcription and smFISH data using Bayesian (PyMC) inference. Four degradation models are compared: spatial ODE, null (constant), age-dependent (Gaussian random walk), and biphasic (poly-A).

Setup

# Create and activate the Python environment
uv sync

Pipelines

Main analysis (Snakefile)

Processes transcription data, bins mRNA counts, runs MCMC inference for all four models, and produces posterior predictive checks and model comparisons.

# Dry-run (preview)
snakemake --cores 1 --dry-run

# Local run
snakemake --cores 4

# SLURM (multicore partition, 36 cores)
sbatch slurm.sh

AP-shift sensitivity (Snakefile_shift_sensitivity)

Repeats the full analysis with the stripe AP window shifted ±half a bin width in each direction, to verify that spatial misalignment between the Berrocal transcription data and smFISH imaging does not bias the inferred degradation rates.

Prerequisite: the main Snakefile must have completed stripe identification first (results_{max_time}/config.yaml.updated must exist).

# Dry-run (preview)
snakemake --snakefile Snakefile_shift_sensitivity --cores 1 --dry-run

# Local run
snakemake --snakefile Snakefile_shift_sensitivity --cores 4

Outputs are written to shiftLeft/ and shiftRight/ directories, keeping them separate from the main results.