A gene regulatory network inference model based on pseudo-siamese network.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) are crucial for understanding organism development and gene interactions.
• Inferring GRNs from single-cell RNA sequencing (scRNA-seq) data is challenging due to noise and data sparsity.
• Existing machine-learning models struggle with the complexity of high-dimensional scRNA-seq data for large-scale GRN inference.

Purpose of the Study:

• To develop a robust framework for inferring large-scale gene regulatory networks (GRNs) from time-series expression data.
• To address the limitations of current methods in handling noisy and high-dimensional scRNA-seq data.
• To elucidate the temporal and spatial features of transcription factors (TFs) and their target genes.

Main Methods:

• Proposed a multilevel, multi-structure framework named pseudo-Siamese GRN (PSGRN).
• Utilized gated recurrent units to capture temporal features and DenseNet for spatial features of TF-target matrices.
• Applied a sigmoid function for interaction evaluation and validated on maize datasets.

Main Results:

• The PSGRN framework demonstrated superior performance compared to state-of-the-art GRN inference methods.
• Successfully inferred GRNs from scRNA-seq data, capturing temporal and spatial gene expression dynamics.
• The model exhibited robustness and generalization capabilities in GRN inference.

Conclusions:

• The PSGRN framework offers a powerful new approach for large-scale GRN inference from time-series scRNA-seq data.
• Provides a theoretical foundation for associating maize genotype-phenotype relationships, aiding breeding efforts.
• Highlights the potential of deep learning for unraveling complex gene regulatory mechanisms.