SCHNEL: scalable clustering of high dimensional single-cell data
Tamim Abdelaal1,2, Paul de Raadt2, Boudewijn P F Lelieveldt1,2
1Delft Bioinformatics Lab, Delft University of Technology, 2628 XE Delft, The Netherlands.
Bioinformatics (Oxford, England)
|December 31, 2020
Summary
SCHNEL is a new, automated tool that scales graph clustering for single-cell data analysis. It efficiently identifies cell populations in millions of cells, outperforming existing methods.
Area of Science:
- Computational biology
- Bioinformatics
- Machine learning
Background:
- Single-cell data analysis requires robust cell population identification through clustering.
- Existing clustering tools struggle with scalability for large single-cell datasets (millions of cells).
- Current methods often lack full automation or require manual parameter estimation.
Purpose of the Study:
- To develop a scalable, reliable, and automated clustering tool for high-dimensional single-cell data.
- To address the limitations of existing clustering methods in handling large datasets.
- To provide a generalizable clustering solution applicable to various single-cell data types and machine learning benchmarks.
Main Methods:
- Developed SCHNEL, a novel tool that transforms high-dimensional data into a hierarchical representation.
- Integrated hierarchical data representation with graph clustering to enhance scalability.
- Applied SCHNEL to cytometry, single-cell RNA sequencing, and MNIST benchmark datasets.
Main Results:
- SCHNEL demonstrated scalability to datasets with 3.5 and 17.2 million cells.
- Outperformed three popular cytometry data clustering tools across seven datasets.
- Achieved meaningful clustering results within practical timeframes.
- Validated SCHNEL's generalizability on single-cell RNA sequencing and MNIST datasets.
Conclusions:
- SCHNEL offers a scalable, automated, and reliable solution for single-cell data clustering.
- The hierarchical data transformation approach enables efficient graph clustering on massive datasets.
- SCHNEL advances the field of single-cell data analysis by enabling robust cell population identification at scale.
