Sceptic: pseudotime analysis for time-series single-cell sequencing and imaging data.

Area of Science:

• Computational Biology
• Single-cell Genomics
• Machine Learning

Background:

• Single-cell RNA sequencing (scRNA-seq) enables studying cell state dynamics.
• Existing pseudotime methods struggle with accuracy and data type generalization.
• Challenges persist in analyzing temporal cell trajectories across diverse single-cell data.

Purpose of the Study:

• To introduce Sceptic, a novel supervised model for pseudotime analysis.
• To enhance the predictive accuracy of temporal ordering in single-cell data.
• To demonstrate Sceptic's versatility across multiple single-cell data modalities.

Main Methods:

• Developed Sceptic, a support vector machine (SVM) based model.
• Employed supervised learning for pseudotime embedding construction.
• Validated Sceptic on time-series scRNA-seq, scATAC-seq, and microscopy image data.

Main Results:

• Sceptic significantly outperforms state-of-the-art pseudotime prediction methods.
• Achieved superior accuracy in determining temporal cell states.
• Demonstrated successful application to diverse single-cell data types, including imaging data.

Conclusions:

• Sceptic offers a robust and accurate solution for single-cell pseudotime analysis.
• The model's generalizability expands its utility beyond scRNA-seq.
• Sceptic advances the field of computational single-cell data interpretation.