A Multiresolution Breast Cancer CIBERSORTx Resource Validated for Accuracy, Interpretive Limits, and Biological and...

Area of Science:

• Oncology
• Bioinformatics
• Genomics

Background:

• Accurate deconvolution of bulk transcriptomes is crucial for understanding the breast cancer tumor microenvironment (TME).
• Existing reference matrices often fail to capture the full spectrum of tumor-specific cellular diversity.
• This limitation hinders comprehensive analysis of the complex cellular composition within tumors.

Purpose of the Study:

• To develop and validate breast cancer-specific, multiresolution CIBERSORTx signature matrices using single-cell RNA sequencing data.
• To systematically evaluate the analytical performance and interpretability of these novel matrices.
• To provide a reliable resource for TME deconvolution in breast cancer research.

Main Methods:

• Construction of major-, minor-, and subset-level signature matrices.
• Assessment of matrix performance using pseudo-bulk mixtures and pure cell profiles.
• Evaluation of biological and clinical coherence in TCGA-BRCA and I-SPY2 cohorts.

Main Results:

• All developed matrices accurately reconstructed pseudo-bulk compositions, with performance decreasing at finer resolutions.
• Increased granularity led to higher spillover, primarily within related cell lineages.
• Inferred cell populations generally showed biologically coherent associations, though some fine-resolution subsets displayed non-canonical patterns.
• Plasmablasts and myeloid populations correlated with pathological complete response, while fibroblastic populations showed negative associations in the I-SPY2 cohort.

Conclusions:

• The developed breast cancer-specific matrices offer a validated and interpretable resource for TME deconvolution.
• The study clarifies the performance characteristics and limitations of multiresolution deconvolution.
• Findings highlight the importance of cross-lineage transcriptional contrast for accurate deconvolution.