Multi-resolution characterization of molecular taxonomies in bulk and single-cell transcriptomics data

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• High-throughput genomics assays are increasingly used in large-scale biomedical research.
• These studies often require data-driven discovery of molecular subtypes and relationships.
• Existing methods may not be universally applicable across diverse omics data types.

Purpose of the Study:

• To introduce K2Taxonomer, a novel unsupervised recursive partitioning algorithm and R package.
• To enable the identification of robust subgroups with a taxonomy-like structure.
• To provide a versatile tool for analyzing various omics data, including bulk and single-cell transcriptomics.

Main Methods:

• Developed K2Taxonomer, an unsupervised recursive partitioning algorithm utilizing ensemble learning.
• Designed K2Taxonomer to accommodate different data paradigms (bulk and single-cell transcriptomics, other omics).
• Validated K2Taxonomer's performance on simulated and human tissue data.

Main Results:

• K2Taxonomer successfully identified known relationships in diverse datasets.
• Demonstrated the algorithm's capability to discover robust subgroups across different data types.
• Applied K2Taxonomer to breast cancer tumor-infiltrating lymphocyte (TIL) single-cell profiles.

Conclusions:

• K2Taxonomer is a powerful tool for unsupervised subgroup discovery in large-scale omics data.
• Identified co-expression of translational machinery genes in T cell subtypes within breast cancer TILs.
• This program is associated with better prognosis in breast cancer, highlighting K2Taxonomer's clinical relevance.