bioRxivBioinformaticsPreprint
A novel metric reveals previously unrecognized distortion in dimensionality reduction of scRNA-Seq data
Timothy HamiltoniD, Breanne SpartaiD, Shamus M. CooleyiD, Samuel D. AragonesiD, J. Christian J. RayiD, Eric J. DeedsiD
Abstract summary
Claude Opus 5.5 · xhigh effort · 2 Nov 2026 · from the abstract · CC BY-NC-ND 4.0
Comparing local point neighborhoods before and after dimensionality reduction reveals that standard scRNA-seq pipelines, t-SNE, and UMAP often distort local neighborhoods by more than 95%, altering downstream clustering.
Standard scRNA-seq pipelines reduce high-dimensional gene expression matrices before performing downstream tasks such as cell type clustering and trajectory inference that depend on near-neighbor relationships. Quantifying local neighborhood preservation before and after dimensionality reduction reveals how severely common embeddings alter local topological structure.
Key results
ShownStandard dimensionality reduction steps in scRNA-seq pipelines severely disrupt local point neighborhoods and fail to preserve consistent neighbor sets across pipeline stages.
Comparing local neighborhoods before and after dimensionality reduction showed that local neighborhood distortion in scRNA-seq datasets frequently exceeded 95%, with no consistent set of neighborhoods maintained across steps of the consensus scRNA-seq pipeline. · Abstract
ShownPopular embedding methods such as t-SNE and UMAP distort local structure even on simple simulated data and substantially alter downstream cell type clustering.
Neighborhood comparisons demonstrated substantial local distortion from t-SNE and UMAP on simple simulated datasets and showed that neighborhood distortion strongly affected cell type clustering and other downstream scRNA-seq analyses. · Abstract
Methods, data and reproducibility
Developed a metric that quantifies topological distortion by comparing the local neighborhoods of data points before and after dimensionality reduction across simulated datasets and steps of the consensus scRNA-seq analysis pipeline.
- Simulated datasets and scRNA-seq datasets
- Methods & tools
- t-SNE, UMAP
- Reproducibility gaps
- No code repository, mathematical definition of the neighborhood metric, or specific scRNA-seq dataset accessions are included in the abstract.
Limitations (1)
- The abstract does not specify which biological scRNA-seq datasets, neighborhood sizes, or clustering algorithms were evaluated. · noted in this summary
Read this if you rely on t-SNE, UMAP, or reduced-dimension embeddings for scRNA-seq clustering and trajectory inference and want to assess local neighborhood distortion.
- Citations
- 53
Abstract
High-dimensional data are becoming increasingly common in nearly all areas of science. Developing approaches to analyze these data and understand their meaning is a pressing issue. This is particularly true for single-cell RNA-seq (scRNA-seq), a technique that simultaneously measures the expression of tens of thousands of genes in thousands to millions of single cells. Popular analysis pipelines significantly reduce the dimensionality of the dataset before performing downstream analysis. One problem with this approach is that dimensionality reduction can introduce substantial distortion into the data, particularly by disrupting the local neighborhoods of certain points. Since many scRNA-seq analyses like cell type clustering or trajectory inference rely on these near-neighbor relationships, distortion in this aspect of the data could significantly influence the outcomes of these analyses. Here, we introduce a straightforward approach to quantifying this distortion by comparing the local neighborhoods of points before and after dimensionality reduction. We found that popular techniques like t-SNE and UMAP introduce substantial distortion even for simple simulated data sets. For scRNA-seq data, we found the distortion in local neighborhoods was often greater than 95%, and that there was no consistent set of neighborhoods across the various steps in the consensus scRNA-seq analysis pipeline. We also found that this distortion had profound impacts on the outcomes of cell type clustering and other downstream analyses. Our findings suggest that caution must be applied when interpreting results in terms of 2-D visualizations produced by tools like UMAP, and that there is a critical need for new dimensionality reduction tools that more effectively preserve the local topological structure of the data.
- assay: scRNA-seq
- research focus: benchmarking
- research focus: dimensionality reduction
- approach: statistical model
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- A quantitative framework for evaluating single-cell data structure preservation by dimensionality reduction techniques
Cody N. Heiser, Ken S. Lau·Cell ReportsPeer-reviewed·May 1, 2020
- From Pairwise Distances to Neighborhood Preservation: Benchmarking Dimensionality Reduction Algorithms for CyTOF, scRNA-seq, and CITE-seq
Polina Bombina, Zachary B. Abrams, … Kevin R. Coombes·bioRxiv·Apr 28, 2025
- Accuracy, Robustness and Scalability of Dimensionality Reduction Methods for Single Cell RNAseq Analysis
Shiquan Sun, Jiaqiang Zhu, … Xiang Zhou·Genome biologyPeer-reviewed·Dec 1, 2019
- Comparison of Principal Component Analysis and t-Stochastic Neighbor Embedding with Distance Metric Modifications for Single-cell RNA-sequencing Data Analysis
Haejoon Kwon, Jean Fan, Peter V. Kharchenko·bioRxiv·Jan 29, 2017
- Hubness reduction improves clustering and trajectory inference in single-cell transcriptomic data
Elise Amblard, Jonathan Bac, … Andrei Y. Zinovyev·BioinformaticsPeer-reviewed·Nov 17, 2021
- Supervised Application of Internal Validation Measures to Benchmark Dimensionality Reduction Methods in scRNA-seq Data
Forrest C. Koch, Gavin J. Sutton, … Fatemeh Vafaee·Briefings in BioinformaticsPeer-reviewed·Jul 19, 2021