TransGRN: A Transfer Learning-Based Framework for Inferring Gene Regulatory Networks Across Cell Lines

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Gene regulatory networks (GRNs) are crucial for cellular function.
• Single-cell RNA sequencing (scRNA-seq) enables high-resolution GRN inference.
• Existing methods struggle with limited data (few-shot learning) due to reliance on prior information.

Purpose of the Study:

• To develop a novel computational method for inferring GRNs across cell lines, particularly in data-scarce scenarios.
• To overcome the limitations of existing GRN inference approaches in few-shot settings.
• To improve the accuracy and applicability of GRN inference using transfer learning.

Main Methods:

• Proposed TransGRN, a transfer learning-based method for GRN inference.
• Utilized a cross-cell-line pre-training strategy with scRNA-seq data from multiple sources.
• Integrated biological knowledge from large language models.
• Developed a regulatory interaction extraction module combining gene expression and semantic information.

Main Results:

• TransGRN demonstrated state-of-the-art performance on benchmark tests.
• Achieved superior results in few-shot GRN inference tasks.
• Successfully transferred generalizable gene-gene regulatory patterns across cell lines.

Conclusions:

• TransGRN effectively addresses the challenge of GRN inference with limited data.
• The method enhances GRN analysis by leveraging cross-cell-line learning and large language models.
• Offers a powerful tool for understanding cellular mechanisms in diverse cell types.