DeepIMAGER: Deeply Analyzing Gene Regulatory Networks from scRNA-seq Data.

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Gene regulatory networks (GRNs) are crucial for cellular processes, but inferring cell-specific GRNs is challenging.
• Existing computational methods using RNA sequencing data often yield low precision and high false positives.

Purpose of the Study:

• To develop an advanced computational tool, DeepIMAGER, for accurate inference of cell-specific GRNs.
• To leverage deep learning and data integration for improved GRN inference.

Main Methods:

• DeepIMAGER uses a supervised deep learning approach, transforming gene pair co-expression into image-like data.
• The model is trained on single-cell RNA sequencing and ChIP-sequencing data, incorporating transcription factor binding information.
• It is validated against ten popular GRN inference tools across six cell lines.

Main Results:

• DeepIMAGER demonstrates superior performance and robustness against data dropout compared to existing tools.
• Application to multiple myeloma datasets identified potential GRNs for key transcription factors in dendritic cells.
• The tool accurately decodes GRNs from single-cell RNA sequencing data.

Conclusions:

• DeepIMAGER represents a significant advancement in inferring cell-specific GRNs.
• This tool offers high accuracy and robustness, making it valuable for studying complex regulatory networks.
• DeepIMAGER facilitates a deeper understanding of molecular mechanisms in diverse cell types.