CDState Resolves Malignant Cell Heterogeneity from Bulk Tumor RNA-Sequencing Data

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Intratumor transcriptional heterogeneity (ITTH) complicates cancer treatment and response.
• Single-cell RNA sequencing (scRNA-seq) can resolve ITTH but is costly and technically demanding.
• Bulk RNA sequencing (bulk RNA-seq) is scalable but requires computational deconvolution for cell state analysis.

Purpose of the Study:

• To develop an unsupervised computational method for inferring malignant cell states from bulk RNA-seq data.
• To address limitations of existing supervised and unsupervised deconvolution methods for cancer ITTH.
• To establish a framework for investigating malignant ITTH in large-scale cancer datasets.

Main Methods:

• Developed CDState, an unsupervised deconvolution method using nonnegative matrix factorization with sum-to-one constraint and cosine-similarity optimization.
• Validated CDState using pseudobulk scRNA-seq datasets across five cancer types.
• Applied CDState to 33 cancer types from The Cancer Genome Atlas (TCGA).

Main Results:

• CDState accurately estimated cell-specific gene expression and proportions, outperforming existing unsupervised methods.
• Identified recurrent gene programs (e.g., EMT, MYC targets, oxidative phosphorylation) driving malignant ITTH.
• Linked malignant cell state proportions to clinical features like patient survival and therapeutic response.
• Discovered potential genetic drivers (e.g., TP53, KRAS, PIK3CA) of malignant ITTH.

Conclusions:

• CDState effectively characterizes malignant cell states from bulk RNA-seq data.
• The method provides a framework for studying malignant ITTH in large cancer atlases.
• Understanding ITTH is crucial for improving cancer treatment strategies and patient outcomes.