Finding differentially expressed genes between cell fates predicted by image-based deep learning.

Area of Science:

• Cell Biology
• Genomics
• Computational Biology

Background:

• Identifying early gene expression changes is crucial for understanding cell differentiation and disease.
• Single-cell RNA sequencing (RNA-seq) is powerful for gene expression analysis but struggles with early cell fate determination due to cell disruption.
• Deep learning excels at predicting cell fates from images, offering a complementary approach.

Purpose of the Study:

• To develop and validate an integrated approach combining deep learning-based cell fate prediction from images with single-cell whole-transcriptome analysis.
• To identify differentially expressed genes (DEGs) associated with distinct cell fates, particularly in early stages.
• To apply the method to distinguish between cells fated to survive and die following heat stress.

Main Methods:

• Applied heat stress to a mammalian cell line to induce cell death.
• Utilized time-lapse imaging and developed deep learning models for image-based cell fate prediction (survival vs. death).
• Performed single-cell RNA sequencing on cells post-imaging and compared transcriptomes between predicted fates to detect DEGs.

Main Results:

• Successfully identified DEGs between cells predicted to die and survive.
• Demonstrated the utility of the integrated approach in detecting significant gene expression differences even when transcriptomic profiles lacked clear clustering.
• Validated the method's ability to find fate-related genes in early cell fate determination.

Conclusions:

• The developed approach effectively integrates image-based deep learning with single-cell transcriptomics to identify early cell fate-related genes.
• This method advances the understanding of cell fate regulation and facilitates the discovery of novel molecular markers.
• It offers a powerful tool for hypothesis-free discovery in cell biology and disease research.