GRNUlar: A Deep Learning Framework for Recovering Single-Cell Gene Regulatory Networks.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) are crucial for understanding cellular processes.
• Inferring GRNs from single-cell RNA sequencing (scRNA-Seq) data presents significant computational challenges.
• Existing methods often struggle to capture the complexity and sparsity of GRNs.

Purpose of the Study:

• To develop a novel deep learning framework, GRNUlar, for supervised GRN inference from scRNA-Seq data.
• To enhance the accuracy and efficiency of GRN inference.
• To leverage multitask learning and unrolled algorithms for improved GRN reconstruction.

Main Methods:

• Developed GRNUlar, a deep learning framework integrating multitask learning and an unrolled algorithm.
• Utilized synthetic scRNA-Seq data simulators for supervised training.
• Employed neural networks to model complex transcription factor-gene dependencies.

Main Results:

• GRNUlar demonstrated superior performance compared to state-of-the-art methods.
• The framework successfully inferred GRNs from both synthetic and real scRNA-Seq datasets.
• Validated the effectiveness of using expression data simulators for supervised GRN inference.

Conclusions:

• GRNUlar offers a powerful new approach for GRN inference from scRNA-Seq data.
• The study highlights the potential of deep learning and synthetic data in advancing systems biology.
• GRNUlar advances the field of computational genomics and gene regulatory network analysis.