scURL: Uncertainty-Sharpened Representation Learning for Single-cell Multi-omics Clustering.

scURL achieves superior clustering accuracy and robustness in single-cell multi-omics data, outperforming existing methods across ten datasets with improved biological relevance.

• Why it matters: Accurately integrating diverse omics layers at the single-cell level is challenging due to variable cell-wise reliability and cross-omics heterogeneity, which can bias data interpretation and hinder biological insights.
• What they did: The framework introduces representation uncertainty and uncertainty-aware fusion, supported by a multi-granular calibration module that refines omics-specific representations before fusion, utilizing ten datasets for validation.
• The result: scURL's approach enables more reliable cell population identification, maintains robustness under dropout noise, and facilitates biologically meaningful discoveries, advancing single-cell multi-omics analysis.