A universal approach for integrating super large-scale single-cell transcriptomes by exploring gene rankings

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) generates vast amounts of expression data.
• Existing tools struggle to integrate exponentially accumulating, large-scale single-cell datasets.
• Efficient integration methods are crucial for advancing translational single-cell research.

Purpose of the Study:

• To present iSEEEK, a universal approach for integrating super large-scale single-cell expression data.
• To demonstrate the efficiency and applicability of iSEEEK across diverse datasets.
• To enable knowledge transfer from large datasets to new, unannotated data.

Main Methods:

• Developed iSEEEK, a novel computational approach utilizing expression rankings of top-expressing genes.
• Trained and validated iSEEEK on a dataset comprising 11.9 million single cells.
• Evaluated iSEEEK's performance on five heterogeneous human and mouse datasets for downstream tasks.

Main Results:

• iSEEEK achieved effective integration of super large-scale single-cell transcriptomes.
• Demonstrated robust clustering performance, aligning well with annotated cell labels.
• Showcased successful knowledge transfer capabilities to new datasets and enabled identification of cell-type-specific gene-gene interaction networks.

Conclusions:

• iSEEEK provides a simple yet effective method for integrating massive single-cell expression datasets.
• The approach facilitates accurate cell type identification and understanding of gene interactions.
• iSEEEK has the potential to significantly advance bench-to-bedside translational single-cell research.