DestinyNet: A deep-learning framework for cell-fate analysis from lineage-tracing single-cell RNA sequencing data.

Area of Science:

• Developmental Biology
• Computational Biology
• Genomics

Background:

• Understanding cell development, lineage commitment, and differentiation is crucial in biology.
• Existing cell-fate analysis tools for lineage-tracing single-cell RNA sequencing (LT-scSeq) data lack comprehensive frameworks for accuracy, robustness, and scalability.

Purpose of the Study:

• To introduce DestinyNet, a novel multi-task deep-learning framework for advanced cell-fate analysis.
• To address key challenges in LT-scSeq data, including fate clustering, fate flow visualization, and early cell-fate bias prediction.

Main Methods:

• DestinyNet employs end-to-end cell representation learning using cell-relation triplets.
• The framework integrates fate and cell-type information for enhanced clustering.
• It visualizes dynamic pseudotime trajectories incorporating fate information for fate flow analysis.

Main Results:

• DestinyNet demonstrates robustness across diverse LT-scSeq data types (static, cumulative, dynamic barcoding; single/multiple time points).
• Experiments on hematopoiesis differentiation and fibroblast reprogramming datasets confirm its effectiveness.
• The framework successfully performs multiple cell-fate analysis tasks, including prediction of early cell-fate biases.

Conclusions:

• DestinyNet provides a comprehensive and effective solution for cell-fate analysis using LT-scSeq data.
• The framework enhances the understanding of cell development dynamics and disease progression.
• DestinyNet offers a scalable and robust approach for analyzing complex biological systems.