dynDeepDRIM: a dynamic deep learning model to infer direct regulatory interactions using time-course single-cell gene...

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Time-course single-cell RNA sequencing (scRNA-seq) data reveal dynamic gene expression changes crucial for understanding cell-type-specific gene regulatory networks (GRNs).
• Existing tools often analyze bulk data or static scRNA-seq data, limiting their ability to capture dynamic GRN information.
• There is a need for methods that effectively leverage time-course scRNA-seq data for accurate GRN reconstruction.

Purpose of the Study:

• To introduce dynDeepDRIM, a novel deep learning model for reconstructing GRNs from time-course scRNA-seq data.
• To demonstrate dynDeepDRIM's capability in modeling gene expression dynamics and removing spurious interactions.
• To evaluate dynDeepDRIM's performance against existing methods in inferring transcription factor (TF)-gene interactions and annotating gene functions.

Main Methods:

• Developed dynDeepDRIM, a deep learning model that represents gene pair expression as images.
• Utilized TF-gene pair images and neighboring gene expression context to reconstruct GRNs.
• Modeled gene expression dynamics using high-dimensional tensors and incorporated neighborhood context to eliminate transitive interactions.

Main Results:

• dynDeepDRIM significantly outperformed six other GRN reconstruction methods on simulated and four real time-course scRNA-seq datasets.
• The model effectively inferred TF-gene interactions and reduced false positives.
• dynDeepDRIM demonstrated superior performance in gene function annotation compared to other tools by considering neighbor gene information.

Conclusions:

• dynDeepDRIM provides a powerful and accurate approach for reconstructing GRNs from time-course scRNA-seq data.
• The method's ability to model dynamics and leverage neighborhood context enhances GRN inference and gene function prediction.
• dynDeepDRIM represents a significant advancement in analyzing dynamic cellular processes using scRNA-seq data.