Integrating temporal single-cell gene expression modalities for trajectory inference and disease prediction

Area of Science:

• Single-cell genomics
• Computational biology
• Systems biology

Background:

• Current single-cell analysis relies on static gene expression, limiting understanding of dynamic cellular processes.
• Capturing temporal changes is vital for interpreting cell cycle, development, and disease progression.
• RNA velocity offers insights into transcriptional dynamics but its predictive potential is underexplored.

Purpose of the Study:

• To benchmark integration approaches for temporal single-cell sequencing data.
• To evaluate the utility of integrated modalities for predicting cellular dynamics and phenotypes.
• To provide practical recommendations for integrating gene expression modalities.

Main Methods:

• Benchmarking ten integration approaches across ten diverse datasets.
• Utilizing datasets from various biological contexts, sequencing technologies, and species.
• Comparing integrated data performance against static measurements for trajectory inference and classification.

Main Results:

• Integrated temporal data significantly improves inference of biological trajectories.
• Enhanced performance in classifying cells based on perturbation and disease states.
• Simple concatenation of spliced and unspliced RNA molecules is a robust and efficient method for classification tasks.

Conclusions:

• Integrated temporal gene expression modalities can predict cellular trajectories and disease phenotypes.
• This study offers practical guidance for task-specific integration of single-cell data.
• The findings highlight the value of leveraging temporal information for predictive single-cell modeling.