Trajectory-informed gene feature selection in single-cell analysis with SEEK-VFI

Area of Science:

• Computational Biology
• Machine Learning
• Genomics

Background:

• Identifying highly variable genes is crucial for single-cell trajectory inference.
• Standard methods struggle to distinguish trajectory-relevant genes from noise in continuous cell development.
• This limitation impacts the accuracy of inferring cellular progression.

Purpose of the Study:

• To develop an improved method for prioritizing trajectory-relevant genes in single-cell data.
• To address the limitations of existing methods in continuous developmental trajectories.
• To enhance the preprocessing step for trajectory inference algorithms.

Main Methods:

• Introduced SEEK-VFI (Spectral Ensembling of topic models with Eigenscore for K-agnostic Variable Feature Identification).
• Utilized an ensemble topic modeling approach with eigenscore for gene prioritization.
• Applied machine learning for trajectory inference preprocessing.

Main Results:

• SEEK-VFI outperforms existing methods in identifying trajectory-relevant genes.
• The method successfully differentiates informative genes from uninformative ones.
• Identified key genes that significantly improve trajectory topology reconstruction and visualization.

Conclusions:

• SEEK-VFI provides a robust solution for gene prioritization in single-cell trajectory inference.
• The identified genes enhance downstream trajectory analyses and biological interpretation.
• This approach advances the field of computational biology and single-cell data analysis.