Label-aware distance mitigates temporal and spatial variability for clustering and visualization of single-cell gene...

Area of Science:

• Computational Biology
• Genomics
• Data Science

Background:

• Single-cell gene expression analysis relies on clustering and visualization.
• Euclidean distance in current methods is suboptimal for handling batch effects.
• Batch effects introduce variability, obscuring true cell identities in datasets.

Purpose of the Study:

• Introduce Label-Aware Distance (LAD) to address limitations in single-cell data analysis.
• Improve cell embedding and overcome batch effect challenges.
• Enhance biological discovery through better data interpretation.

Main Methods:

• Developed Label-Aware Distance (LAD), a novel metric.
• Utilized temporal/spatial locality to control for batch effects.
• Validated LAD on simulated and real-world datasets (mouse retina, lung, COVID-19).

Main Results:

• LAD demonstrates superior cell embedding compared to state-of-the-art batch correction methods on longitudinal datasets.
• The metric effectively controls for batch effects in single-cell gene expression data.
• LAD aids in understanding biological processes like mouse development and COVID-19 progression.

Conclusions:

• Label-Aware Distance (LAD) offers a robust solution for batch effect correction in single-cell genomics.
• LAD improves the accuracy of clustering and visualization for biological insights.
• The method facilitates the integration of multi-sample data for enhanced discovery.