BioBERT based text mining for incorporating prior knowledge in the inference of genetic network models

Area of Science:

• Systems Biology
• Computational Biology
• Bioinformatics

Background:

• Gene Regulatory Network (GRN) reconstruction is crucial for understanding cellular processes and disease mechanisms.
• S-system models, using non-linear differential equations, capture complex biological dynamics but face computational challenges with increasing network size.
• Prior biological knowledge integration can mitigate computational demands and improve GRN inference accuracy.

Purpose of the Study:

• To introduce PRESS (Prior Knowledge Enhanced S-system model), an automated framework for accurate GRN reconstruction.
• To enhance the speed and accuracy of GRN inference by incorporating prior knowledge extracted from scientific literature.
• To address the computational limitations of traditional S-system models in large-scale network analysis.

Main Methods:

• Developed PRESS, an integrated model combining automated prior knowledge extraction with an S-system GRN reconstruction approach.
• Utilized a BioBERT-based Gene Interaction Extraction Framework for targeted genetic relation extraction and regulatory gene prediction.
• Implemented a novel fitness evaluation in the optimization algorithm to limit regulatory genes, mimicking real GRNs.

Main Results:

• Demonstrated substantial reductions in computational cost for GRN reconstruction.
• Achieved significant improvements in prediction accuracy compared to existing methods.
• Validated the effectiveness of PRESS using Escherichia coli subnetworks and the benchmark SOS dataset.

Conclusions:

• PRESS offers a novel and effective approach to automated GRN reconstruction by integrating prior biological knowledge.
• The method significantly enhances computational efficiency and prediction accuracy, overcoming limitations of traditional S-system models.
• PRESS represents a significant advancement in computational biology for analyzing complex gene regulatory systems.