SIGNET: single-cell RNA-seq-based gene regulatory network prediction using multiple-layer perceptron bagging.

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables gene regulatory interaction analysis.
• Complex, nonlinear gene interactions are challenging for traditional linear models.
• Transcription factors (TFs) are key regulators of gene expression.

Purpose of the Study:

• To present SIGNET, a deep learning framework for inferring gene regulatory networks from scRNA-seq data.
• To capture complex, nonlinear gene regulatory relationships.
• To improve the identification of regulatory interactions, particularly in rare cell populations.

Main Methods:

• Developed SIGNET, a deep learning framework utilizing TF expression as predictors for target gene expression.
• Applied SIGNET to diverse real and simulated scRNA-seq datasets.
• Validated inferred interactions against ChIP-seq data.

Main Results:

• SIGNET effectively captures complex gene regulatory relationships.
• The framework demonstrates higher sensitivity in detecting ChIP-seq validated interactions compared to existing methods.
• SIGNET shows particular efficacy in identifying interactions within rare cell types.

Conclusions:

• SIGNET is a powerful tool for deciphering gene regulatory networks from scRNA-seq data.
• Its ability to capture nonlinear interactions enhances downstream analyses like cell clustering and network inference.
• SIGNET aids in identifying key regulatory modules driving biological processes.