PROFET Predicts Continuous Gene Expression Dynamics from scRNA-seq Data to Elucidate Heterogeneity of Cancer...

Area of Science:

• Computational Biology
• Genomics
• Systems Biology

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution snapshots of cellular gene expression.
• Current scRNA-seq methods cannot capture continuous gene expression dynamics over time.
• Understanding dynamic gene expression is crucial for deciphering cell state transitions and responses to stimuli.

Purpose of the Study:

• To develop a computational framework, PROFET (Particle-based Reconstruction Of generative Force-matched Expression Trajectories), for reconstructing continuous single-cell gene expression trajectories from sparse scRNA-seq data.
• To apply PROFET to investigate cellular heterogeneity and dynamic responses to palbociclib treatment in breast cancer.

Main Methods:

• PROFET utilizes a novel Lipschitz-regularized gradient flow approach to generate particle flows between time-stamped scRNA-seq samples.
• A global vector field for trajectory reconstruction is learned using neural force-matching.
• The framework was validated using synthetic data and mouse/human in vitro datasets.

Main Results:

• PROFET successfully reconstructs nonlinear gene expression trajectories from static scRNA-seq data.
• Application to palbociclib treatment in breast cancer identified a subpopulation of cells with significant phenotypic shifts.
• Unique surface markers enriched in treatment-responsive subpopulations were identified.

Conclusions:

• PROFET enables the inference of continuous single-cell expression trajectories from static data.
• This computational tool is valuable for dissecting cell state heterogeneity during treatment responses.
• PROFET advances the understanding of dynamic cellular processes in complex biological systems.