Dynamic Modelling and Sensitivity Analysis of the Homologous Recombination DNA Repair Network

This repository contains a computational systems-biology project focused on the homologous recombination (HR) DNA repair pathway. The workflow combines graph-based network analysis, ordinary differential equation (ODE) modelling, and sensitivity analysis to identify the molecular mechanisms that most strongly regulate repair progression.

Project objectives

• Reconstruct and analyse a protein-interaction network for homologous recombination.
• Identify structurally important proteins using graph centrality metrics.
• Represent the main HR repair stages with a mechanistic ODE model.
• Refine the RPA-to-RAD51 exchange and RAD51 filament-formation steps.
• Evaluate parameter influence using local sensitivity analysis, Morris screening, and LHS-PRCC analysis.
• Compare the baseline and expanded HR models.

Biological scope

The model represents the progression from a DNA double-strand break to repaired DNA through the following states:

1. Double-strand break recognition
2. DNA end resection
3. RPA-coated single-stranded DNA
4. RAD51 loading
5. RAD51 filament formation
6. Strand invasion and D-loop formation
7. Late repair progression
8. Repaired DNA

The expanded model includes regulators such as CtIP, EXO1, DNA2, DSS1, RAD54, and FANCM.

Main findings

• The simulated trajectories reproduce a biologically coherent progression from DNA damage to repaired DNA.
• RPA-coated ssDNA and the RPA-to-RAD51 exchange step emerge as major regulatory control points.
• BRCA2-DSS1-mediated RAD51 loading and RAD51 filament assembly show strong sensitivity.
• Refining the RAD51 module increases the apparent influence of RAD51-related mechanisms.
• FANCM-related parameters influence D-loop stabilization and downstream repair progression.
• Because most kinetic constants are not experimentally calibrated, the model is exploratory and hypothesis-generating rather than quantitatively predictive.

Installation

Create and activate a virtual environment, then install the required packages:

python -m venv .venv

Windows PowerShell:

.venv\Scripts\Activate.ps1
pip install -r requirements.txt

macOS/Linux:

source .venv/bin/activate
pip install -r requirements.txt

Usage

Network analysis

python src/network_analysis.py

This script reads the interaction table, builds a directed graph, calculates centrality metrics, and saves the results in results/network_centrality.csv.

ODE simulation

python src/ode_model.py

This script simulates the main mechanistic HR repair states and saves the trajectories and figure.

Sensitivity analysis

python src/sensitivity_analysis.py

This script performs a simple local sensitivity analysis on the ODE model and saves a ranked parameter table and heatmap.

Jupyter notebook

jupyter notebook notebooks/homologous_recombination_model.ipynb

Input data

The interaction dataset contains three tab-separated columns:

• NODE: source protein
• INTERACTION: interaction type
• NODE: target protein

Interaction types include interacts with, controls state change of, controls expression of, and in catalysis with.

Limitations

The numerical values used in the demonstration ODE model are illustrative. They should not be interpreted as experimentally calibrated in-vivo kinetic constants or concentrations.

Author

Ania Rotondi

Citation and references

The accompanying report contains the complete bibliography and biological discussion. The project draws on work concerning homologous recombination mechanisms, BRCA2-DSS1-dependent RAD51 loading, dynamic DNA-repair modelling, and Pathway Commons interaction data.