A novel single-cell based method for breast cancer prognosis

Area of Science:

• Computational biology
• Genomics
• Oncology

Background:

• Breast cancer prognosis is complex due to disease heterogeneity.
• Existing computational methods using bulk RNA-seq data have limitations in performance and biological relevance.
• Intra-tumor heterogeneity is often overlooked in current prognosis models.

Purpose of the Study:

• To develop a novel computational method, scPrognosis, for improved breast cancer prognosis using single-cell RNA sequencing (scRNA-seq) data.
• To leverage the Epithelial-to-Mesenchymal Transition (EMT) biological process for enhanced prognostic accuracy.
• To address the limitations of bulk RNA-seq methods by incorporating cellular-level heterogeneity.

Main Methods:

• scPrognosis infers EMT pseudotime and a dynamic gene co-expression network from scRNA-seq data.
• An integrative model selects key EMT-related genes based on expression variation, differentiation, and network roles.
• Selected gene signatures are used as features to build a prediction model with bulk RNA-seq data for prognosis.

Main Results:

• scPrognosis demonstrates superior performance compared to benchmark breast cancer prognosis methods utilizing bulk RNA-seq data.
• The study identified signature genes with dynamic expression changes during EMT.
• These findings may elucidate the link between EMT and clinical outcomes in breast cancer.

Conclusions:

• scPrognosis offers a significant advancement in breast cancer prognosis by effectively utilizing scRNA-seq data.
• The method successfully integrates intra-tumor heterogeneity into prognostic models.
• scPrognosis is adaptable for analyzing other biological processes within scRNA-seq datasets.