Interpretable Trajectory Inference with Single Cell Linear Adaptive Negative-binomial Expression (scLANE) Testing.

Area of Science:

• Computational Biology
• Single-cell Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) enables the study of gene expression dynamics during biological processes.
• Existing trajectory inference methods often use complex models (e.g., generalized additive models) for differential expression analysis, leading to interpretation challenges.
• Subjective visual inspection of gene expression patterns is frequently used to draw biological conclusions.

Purpose of the Study:

• To develop a novel, interpretable method for trajectory differential expression testing in scRNA-seq data.
• To address the limitations of existing methods in handling nonlinearity and facilitating biological interpretation.
• To provide a robust tool for analyzing complex experimental designs in single-cell studies.

Main Methods:

• Introduction of scLANE testing, utilizing an interpretable generalized linear model framework.
• Incorporation of empirical basis splines to handle nonlinear gene expression changes.
• Extensions using estimating equations and mixed models for complex experimental designs and robust testing.

Main Results:

• scLANE testing demonstrated high accuracy across various simulation scenarios.
• The method was successfully applied to diverse biological datasets.
• scLANE testing provided novel biological insights when integrated with pseudotime and RNA velocity estimation.

Conclusions:

• scLANE testing offers a significant improvement in interpretability and accuracy for trajectory differential expression analysis in scRNA-seq data.
• The method enhances the ability to derive meaningful biological conclusions from single-cell gene expression dynamics.
• scLANE testing is a versatile tool applicable to various trajectory inference outputs and experimental designs.