Integrating pathway knowledge with deep neural networks to reduce the dimensionality in single-cell RNA-seq data

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) reveals cellular heterogeneity, crucial for understanding biology and disease.
• Identifying novel cell types and states is a key application of scRNA-seq data analysis.
• Deep neural networks (DNNs) excel at this but often lack interpretability.

Purpose of the Study:

• To develop an intelligible pathway-driven neural network for scRNA-seq data analysis.
• To provide a biologically meaningful representation of single-cell data.
• To improve cell-type identification and classification.

Main Methods:

• Explored DNNs constrained by prior biological information, specifically signaling pathways.
• Tested biologically-based DNN architectures on human and mouse scRNA-seq datasets.
• Validated model performance in cell-type clustering and database querying scenarios.

Main Results:

• The pathway-driven DNN achieved performance comparable to less interpretable DNNs.
• Demonstrated the ability to visualize and interpret the latent structure of human single-cell datasets.
• Showcased accurate cell-type annotation and clustering.

Conclusions:

• Integrating biological pathways with DNNs offers a powerful, interpretable alternative for scRNA-seq analysis.
• Prior biological knowledge in DNNs reduces network size and enhances interpretability.
• The approach provides a biologically meaningful representation of scRNA-seq data, exemplified in human melanoma cells.