GRN-transformer

GRN-transformer reconstructs cell-type-specific gene regulatory networks (GRNs) by integrating single-cell RNA sequencing (scRNA-seq) with bulk-cell transcriptomic data using an axial transformer and weakly supervised learning.

Key Features:

• Integration of Bulk-Cell Data: Incorporates bulk-cell transcriptomic data and bulk-derived GRNs to augment single-cell GRN prediction accuracy.
• Weakly Supervised Learning Framework: Employs a weakly supervised learning approach to combine information from single-cell and bulk-cell sources for cell-type-specific GRN prediction.
• Axial Transformer Architecture: Uses an axial transformer neural network architecture to model and integrate high-dimensional scRNA-seq and bulk-cell data.
• State-of-the-Art Prediction Accuracy: Experimental validation reported higher prediction accuracy compared to existing supervised and unsupervised GRN inference approaches.

Scientific Applications:

• Cell-Type-Specific GRN Inference: Infers gene regulatory networks at cell-type resolution from integrated single-cell and bulk transcriptomic data.
• Identification of Key Transcription Factors and Regulations: Identifies transcription factors and regulatory interactions associated with disease risk genes, including applications to Alzheimer's disease risk genes.
• Broad Applicability in Genomic Research: Applicable to genomic studies that require integrating scRNA-seq with bulk-cell data for detailed regulatory inference.

Methodology:

Uses an axial transformer within a weakly supervised learning framework to integrate scRNA-seq and bulk-cell transcriptomic data or bulk-derived GRNs to predict cell-type-specific GRNs.