EnsembleRegNet: Interpretable deep learning for transcriptional network inference from single-cell RNA-seq.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Gene regulatory networks (GRNs) are crucial for cellular identity and gene expression.
• Inferring GRN structure from single-cell RNA sequencing (scRNA-seq) data is challenging.
• Existing methods often lack accuracy and interpretability.

Purpose of the Study:

• To develop a robust and interpretable deep learning framework for GRN inference.
• To improve the accuracy of transcription factor (TF)-target gene relationship identification.
• To establish a new benchmark for GRN inference from scRNA-seq data.

Main Methods:

• Developed EnsembleRegNet, a deep learning framework integrating an ensemble encoder-decoder and MLP.
• Utilized Hodges-Lehmann estimator (HLE)-based binarization and case-deletion analysis.
• Incorporated RcisTarget for motif enrichment and AUCell for regulon activity scoring.

Main Results:

• EnsembleRegNet demonstrated superior performance compared to SCENIC and SIGNET on simulated and real scRNA-seq datasets.
• Achieved higher accuracy in GRN inference and improved clustering performance.
• Successfully uncovered cell-type-specific regulatory modules.

Conclusions:

• EnsembleRegNet provides a scalable and biologically grounded framework for transcriptional regulation analysis.
• Offers enhanced robustness and interpretability for GRN inference.
• Shows promise for applications in disease modeling, biomarker discovery, and cellular reprogramming.