HyperCell: Advancing Cell Type Classification with Hyperdimensional Computing.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Data Science

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high-resolution insights into cellular heterogeneity.
• scRNA-seq data presents challenges like high dimensionality, noise, and sparsity.
• Accurate cell type classification is crucial for understanding biological systems.

Purpose of the Study:

• To develop a novel approach for enhancing cell type classification accuracy in scRNA-seq data.
• To leverage hyperdimensional computing for robust analysis of noisy and sparse scRNA-seq datasets.
• To compare the proposed method against established classification techniques.

Main Methods:

• Utilized hyperdimensional computing for scRNA-seq data analysis.
• Employed the QuantHD method for high-dimensional hypervector encoding and iterative training.
• Conducted experiments on diverse datasets using split-by-batch and random split settings.

Main Results:

• The proposed hyperdimensional computing approach demonstrated superior performance in handling noisy scRNA-seq data.
• Outperformed established methods including XGBoost, Seurat reference mapping, and scANVI.
• Validated effectiveness across different dataset splitting strategies.

Conclusions:

• Hyperdimensional computing shows significant potential for advancing single-cell data analysis.
• The developed method offers more accurate cell type annotation.
• This work facilitates deeper insights into cellular dynamics, tissue functions, and disease mechanisms, benefiting biomedical research and personalized medicine.