Quantifying the effect of experimental perturbations at single-cell resolution.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Current single-cell RNA sequencing analysis methods often focus on discrete cell clusters.
• Comparing datasets across multiple conditions using these methods can be limited in capturing continuous transcriptional changes.

Purpose of the Study:

• To develop a novel computational framework for quantifying perturbation effects at the single-cell level.
• To enable a more sensitive and accurate comparison of single-cell RNA sequencing datasets across different experimental conditions.

Main Methods:

• Modeling the transcriptomic space as a manifold.
• Utilizing graph signal processing to estimate the relative likelihood of observing cells under various conditions.
• Implementing vertex frequency clustering for identifying affected cell populations.

Main Results:

• The developed algorithm demonstrates a 57% improvement in accuracy for identifying condition-specific cell enrichments or depletions compared to existing methods.
• Gene signatures derived from the identified cell populations show superior accuracy in ground truth comparisons.

Conclusions:

• This continuous, manifold-based approach provides a more robust method for analyzing single-cell RNA sequencing data under perturbation.
• The algorithm accurately identifies cell populations affected by perturbations and generates reliable gene signatures, advancing comparative transcriptomics.