scTIGER: A Deep-Learning Method for Inferring Gene Regulatory Networks from Case versus Control scRNA-seq Datasets.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Inferring gene regulatory networks (GRNs) is crucial for understanding cellular processes.
• Existing methods for GRN inference from single-cell RNA sequencing (scRNA-seq) data often suffer from high false positive rates.
• No current methods directly utilize paired case-versus-control scRNA-seq datasets for GRN inference.

Purpose of the Study:

• To introduce scTIGER, a novel deep-learning-based method for GRN detection.
• To infer GRNs by analyzing co-differential gene expression profiles in paired scRNA-seq datasets.
• To address limitations of existing GRN inference methods, particularly regarding false positives and paired data utilization.

Main Methods:

• scTIGER utilizes paired scRNA-seq datasets from case-versus-control experiments.
• The method incorporates cell-type-based pseudotiming, an attention-based convolutional neural network, and permutation-based significance testing.
• It infers GRNs by analyzing co-differential relationships within gene modules.

Main Results:

• scTIGER successfully identified dynamic regulatory networks in prostate cancer cells, including key genes like AR, ERG, PTEN, and ATF3.
• The method detected specific regulatory networks in neurons related to fear memory, involving genes such as BDNF, CREB1, and MAPK4.
• scTIGER demonstrated robustness against high levels of dropout noise inherent in scRNA-seq data.

Conclusions:

• scTIGER provides a powerful and accurate approach for inferring gene regulatory networks from paired scRNA-seq data.
• The method enhances the understanding of regulatory mechanisms in various biological contexts, including cancer and neuroscience.
• scTIGER's resilience to data noise makes it a valuable tool for analyzing challenging scRNA-seq datasets.