These notebooks reproduce the implementation and analysis of a biophysical model for ATAC-seq data analysis.

The notebooks are arranged as follows:

• moments.ipynb: Calculate transcriptome and chromatin-state moments and compare to simluated results for a two-gene system. Reproduces Figure S5.
• distinguishability_simulation.ipynb: Fit our biophysical model to simulated results and compare parameter identifiability for different modalities. Reproduces Figure 5.
• noise.ipynb: Illustration of technical noise implementation. Reproduces Figure S4.
• loci_statistics.ipynb: Perform exploratory data analysis of ATAC-seq data. Plot site-site correlations and means. Reproduces panels in Figure 2, and Figures S1-S3.
• BIC_analysis.ipynb: Fit models with peak-peak correlations to ATAC data and compare Bayesian Information Criteria with a simple model without correlations. Reproduces panels in Figures 3 and 4.
• toy_systems.ipynb: Illustrate toy systems which have transcript correlations greater than chromatin site correlations. Reproduces panels in Figure S7.

The following scripts contain helper functions arranged as follows:

• construct_H.py: Create transition matrix for Ising chromatin model.
• simulation.py: Simulate RNA counts and chromatin configurations from model.
• noise.py: Technical noise implementation.
• moments.py: Functions for calculating moments.

We also include a script to show how the six-site loci for ATAC-seq analysis were selected, using one of the datasets as an example.

• loci_selection.py

The observed chromatin accessibility at these loci from the processed ATAC-seq data are included in sub-folders.