aKNNO: single-cell and spatial transcriptomics clustering with an optimized adaptive k-nearest neighbor graph.

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Standard clustering methods for transcriptomics data often fail to detect rare cell types.
• Specialized rare cell detection methods may compromise accuracy for abundant cell populations.

Purpose of the Study:

• To develop a novel computational method, aKNNO, for simultaneous identification of abundant and rare cell types in transcriptomic data.
• To improve cell type identification accuracy in single-cell and spatial transcriptomics.

Main Methods:

• Development of aKNNO, utilizing an adaptive k-nearest neighbor graph with optimization.
• Benchmarking against existing methods on 38 simulated and 20 real-world transcriptomics datasets.

Main Results:

• aKNNO demonstrates superior accuracy in identifying both abundant and rare cell types compared to general and specialized methods.
• The method achieves precise mapping of abundant and rare cells using only gene expression data, outperforming integrative approaches.

Conclusions:

• aKNNO offers a robust solution for simultaneous identification of diverse cell populations in transcriptomics.
• The method enhances the analysis of single-cell and spatial transcriptomics data, particularly for complex biological systems.