GripDL

GripDL predicts gene regulatory interactions from spatial gene expression microscopy images using deep learning to reconstruct gene regulatory networks (GRNs) for studying cellular processes and development.

Key Features:

• Spatial image input: Uses microscopy images of spatial gene expression, specifically Drosophila embryonic gene expression images, as primary data for inference.
• Local pattern emphasis: Prioritizes local pattern consistency within images rather than global image similarity for assessing gene-gene relationships.
• Deep learning representations: Employs deep neural networks to learn representations from spatial expression patterns.
• Integration of TF–gene priors: Incorporates high-confidence transcription factor (TF)–gene regulation knowledge from prior studies as prior information.
• Supervised learning framework: Uses the integrated TF–gene prior knowledge to provide supervision for network inference.
• GRN construction: Constructs gene regulatory networks by combining learned image representations with prior regulatory information.
• Application focus: Targets reconstruction and analysis of GRNs involved in Drosophila eye development.
• Addresses limitations of traditional methods: Aims to reduce false positives and missed connections associated with co-expression and TF–target binding–only approaches.

Scientific Applications:

• GRN reverse engineering: Reverse engineers gene regulatory networks from spatial expression data.
• Spatial expression analysis: Analyzes spatial and temporal patterns of gene activity captured by microscopy.
• Drosophila developmental biology: Investigates gene regulatory interactions relevant to Drosophila embryonic eye development.
• Discovery of novel interactions: Enables identification of previously unrecognized gene–gene interactions.
• Systems biology: Supports studies of cellular processes and developmental regulation through reconstructed GRNs.

Methodology:

Combines analysis of spatial microscopy gene expression images with deep neural networks and integrates high-confidence TF–gene regulation knowledge from prior studies in a supervised learning framework that emphasizes local pattern consistency to infer GRNs in Drosophila embryonic data.