Inferring gene regulatory network via fusing gene expression image and RNA-seq data

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Gene regulatory network (GRN) reconstruction is crucial for understanding cellular mechanisms.
• Traditional methods using RNA-seq data have limitations in capturing comprehensive gene expression information.
• Emerging gene expression image databases offer rich spatial context for GRN inference.

Purpose of the Study:

• To develop novel deep learning models for inferring GRNs from gene expression images and RNA-seq data.
• To evaluate the performance of models utilizing single and combined data modalities.
• To enhance the accuracy and comprehensiveness of GRN reconstruction.

Main Methods:

• A convolutional neural network (SDINet) was developed to extract gene expression information and identify gene interactions from images.
• An RNA-model was built based on SDINet principles for RNA-seq data analysis.
• A fusion network was designed to integrate information from both image and RNA-seq data.

Main Results:

• SDINet achieved notable performance on image data (Acc: 0.7196, F1: 0.7374).
• The RNA-model demonstrated strong results on RNA-seq data (Acc: 0.8962, F1: 0.8950).
• The proposed fusion network integrating both data modalities outperformed single-modality approaches (Acc: 0.9116, F1: 0.9118).

Conclusions:

• Integrating gene expression images and RNA-seq data via a fusion network significantly enhances GRN inference.
• Deep learning approaches, including SDINet, are effective for extracting complex information from diverse biological data.
• The developed methods provide a powerful framework for more accurate and comprehensive GRN reconstruction.