A Workflow for Merging Logical Gene Regulatory Network Models

Overview

This repository provides a semi-automated workflow for merging logic models, and demonstrate its application using a set of Acute Myeloid Leukemia (AML) models.

The workflow is described in the manuscript "LM-Merger: A workflow for merging logical models with an application to gene regulation". (link)

Li LX, Aguilar B, Gennari J, Qin G. LM-Merger: A workflow for merging logical models with an application to gene regulation. bioRxiv. 2024 Jan 1;2024.09.13.612961.

The workflow involves sequential steps:

1. Finding Models: This step involves reviewing existing literature, repositories, and databases to identify models with shared components.
2. Standardizing and Annotating Models: Convert models to SBML-qual format, and annotate gene names using HGNC approved symbols.
3. Reproducing Selected Models: Verifying that selected models replicate the behaviors described in their original resources.
4. Merging Models: Using the provided tool to automatically merge models with different logical combination methods (OR, AND, Inhibitor Wins).
5. Evaluating the Merged Models: Comparing the predictive accuracy and robustness of the merged models against the original models and applying the merged models to new, untested scenarios.

Repository Structure

This repository is organized according to the workflow described in the manuscript:

1. Standardizing & Annotating Models:
   • Standardization: Converting models in text file to SBML-qual format.
   • Annotation: Fetching HGNC gene symbols for input SBML-qual models and updating the gene names after manual verification.
2. Reproducing Selected Models:
   • Reproducibility check for each collected model, including:
     • Bonzanni2013
     • Ikonomi2020
     • Krumsiek2011
     • Palma2021 (As Table S3 in the manuscript)
3. Composing Models:
   • Merge logical models: Automatically merging logical models, including the OR, AND, and Inhibitor Wins methods.
   • Support models in:
     • Text files using a EBNF description as in Boolnet
     • SBML-qual files
4. Evaluating the Merged Model:
   • Functions: Provides some helper functions for evaluating logical models.
   • Contains notebooks for various evaluation tasks:
     • Coverage: Assessing the coverage of AML patients with each mutation profiles using BeatAML, TCGA, AMLSG and cBioPortal data.
     • Stable States Heatmap: Visualizing stable states of the merged models and clustering them with individual models.
       • Stable States Heatmap - Asynchronous update (Fig 2B, S1)
       • Stable States Heatmap - Synchronous update (Fig 3B, S1)
     • Correlation with HSC Expression: Analyzing the correlation of model predictions with hematopoietic stem cell expression data. (Fig 2C-E, Fig S2)
     • Correlation with Clinical Outcomes: Separate notebooks for evaluating correlations with different clinical indicators/datasets:
       • Blast percentages from the BeatAML data
         • Using approach similar to Palma et al. (Fig 3C-D, Fig S4)
         • Using all mutations (Fig 3E-F, Fig S4)
       • Blast percentages from the TCGA data (TableS3)
       • Hazard ratio for death from the AMLSG data

• Data: Contains datasets used for model evaluation.

• Models: Stores models in text files and SBML-qual format.

How to Use This Repository

1. Clone the repository:

   git clone https://github.com/IlyaLab/LogicModelMerger.git
   

2. Install and open the CoLoMoTo Notebook (Optional): Please refer to the usage guide on their website.

3. Run notebooks: Navigate to the relevant directory and open the Jupyter notebooks using the CoLoMoTo notebook or your preferred Jupyter environment.

CoLoMoTo Interactive Notebook

Jupyter notebooks in this repository were conducted using the CoLoMoTo Interactive Notebook with Docker image colomoto/colomoto-docker:2024-03-01.

The CoLoMoTo notebook provides a unified environment to edit, execute, share, and reproduce analyses of qualitative models of biological networks.

For more information, visit CoLoMoTo.