GP_Transcription_Dynamics

Python implementation of the transcriptional regulation model with Gaussian processes using GPflow and TensorFlow probability. The code here reproduces modelling results from the paper Combined modelling of mRNA decay dynamics and single-molecule imaging in the Drosophila embryo uncovers a role for P-bodies in 5' to 3' degradation.

Dependencies

See requirenments.txt for full reproducibility; for lighter version of the code check simulations_python39 with lighter requirements.

TRCD

-- contains the transcriptional regulation model (custom implementation of GPR for stacked time series; transcriptional regulation kernel).

Simulations:

-- simulated_examples.py runs an experiment on simulated data (generates the data, fits trcd model and runs MCMC).

Simulations_python39:

-- Same as simulations, but with a few updates from latest version of the libraries (i.e., gpflow);

-- Minimal requirements compared to the main directory;

-- Runs on Mac M1, but tensorflow needs to be installed accordingly (see intructions for tensorflow installation here);

-- simulated_examples.py runs an experiment on simulated data (generates the data, fits trcd model and runs MCMC).

Experiments:

Contains files for the experiments on real data.

-- filter_genes.py/filter_genes_2rbf.py filtering of the genes (identifying differentially expressed genes);

-- fit_model_filtered_genes.py optimization of the parameters in transcriptional regulation model for genes that passed filtering;

-- mcmc_single_gene.py/mcmc_all_genes.py MCMC for uncertainty quantification using MALA on single gene/complete data set.

Utils:

Contains helper functions for loading the data/fitting the model/running MCMC.

Clustering:

-- source code for clustering time series with GPclust.