Literature-derived, context-aware gene regulatory networks improve biological predictions and mathematical modeling
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Motivation
Complex gene regulatory networks (GRNs) underlie most disease processes, and understanding disease-specific network structures and dynamics is crucial for developing effective treatments. Yet, literature-based analyses of GRNs often treat gene regulations as context-independent interactions, overlooking how their biological relevance can differ depending on the disease type, cell lineage, or experimental condition.
Results
In an attempt to improve on existing methods for leveraging knowledge present in the scientific literature, we developed a framework to assign quantitative, context-dependent weights to gene regulations extracted from literature. We demonstrate that the context-specific GRNs reconstructed with our method can effectively capture disease biology, showing strong correlation with transcriptomics across a wide range of diseases. Furthermore, we show that utilizing contextual information improves accuracy in drug-target prediction tasks. Finally, we showcase the utility of the contextualized GRNs through the automated construction of an ordinary differential equation model of a breast cancer-specific signaling network. The large language model-based framework allows the integration of literature- and experimentally derived information and streamlines the process of assembling a biologically relevant and functional mathematical model. Our findings indicate the importance of considering the context when making biological predictions, and we demonstrate the use of natural language processing tools to effectively mine associations between gene regulations and biological contexts.
Availability and implementation
All reproducibility code is available at https://github.com/okadalabipr/context-dependent-GRNs , along with the automated mathematical model construction package at https://github.com/okadalabipr/BioMathForge . The dataset used in this study is available at Zenodo, DOI: 10.5281/zenodo.16416117.
