Answers are grounded in this paper and cite their sources.
Single-cell metabolomics is a powerful approach for investigating cellular heterogeneity and disease-related metabolic mechanisms. However, the acquired data often suffer from high variability, low signal intensity, and substantial technical noise, rendering data preprocessing particularly critical. As a core preprocessing step, normalization plays a vital role in mitigating technical variability and improving cross-sample comparability. To date, a standardized normalization strategy has not been established in single-cell metabolomics, which may introduce uncertainty into analytical results and compromise research reproducibility. In this study, we systematically evaluated five normalization methods—total ion current normalization, internal standard normalization, maximum normalization, probabilistic quotient normalization (PQN), and vector length normalization—using a quality control (QC) dataset and a single-cell metabolomics dataset. These methods were benchmarked across multiple dimensions, including analytical stability, preservation of biological structure and cell-typing performance. The results showed that PQN achieved well-balanced performance across the evaluation metrics, offering the highest intra-group similarity (aPCC = 0.88–0.90) and superior cell-typing performance (cell Local Inverse Simpson’s Index = 1.37; classification accuracy = 0.94). Within the scope of the analyzed datasets, our findings indicate that PQN effectively suppressed technical noise while preserving biological structure, supporting it as a viable normalization option for single-cell metabolomics data preprocessing.
No takes yet. Share an insight, caveat, or question.
Hao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: