FusedFCR: A Fused Forward Continuation-Ratio model for marker selection along cell-fate trajectories

Area of Science:

• Computational Biology
• Genomics
• Developmental Biology

Background:

• Time-course single-cell RNA sequencing (scRNA-seq) provides snapshots of dynamic biological processes like differentiation.
• Existing pseudotime methods struggle to link identified genes to specific transitions between observed stages.
• There is a need for methods that can identify stage-specific marker genes from time-course scRNA-seq data.

Purpose of the Study:

• To develop a novel statistical model, FusedFCR, for analyzing time-course scRNA-seq data.
• To identify genes associated with specific transitions between ordered developmental stages.
• To provide a method that integrates gene selection with pseudotemporal ordering.

Main Methods:

• Proposed FusedFCR, a regularized forward continuation-ratio model.
• Combined lasso penalty for gene selection with a fusion penalty for adjacent transition similarity.
• Developed transition-specific coefficients for interpretable gene selection and stage-anchored pseudotime.

Main Results:

• FusedFCR accurately recovered gene-effect trajectories in simulations.
• Identified biologically interpretable genes associated with distinct transitions in mouse beta-cell and human trophoblast differentiation.
• Demonstrated competitive or superior stage-classification accuracy compared to alternative methods.

Conclusions:

• FusedFCR effectively identifies stage-specific molecular programs during cellular progression.
• The method complements pseudotemporal ordering by pinpointing when specific gene expression changes occur.
• FusedFCR offers a valuable tool for analyzing dynamic biological processes using time-course scRNA-seq data.