A Pan-Cancer Single-Cell Atlas to Evaluate Tumor Identity, Cell Line Concordance, and Dependency Mapping.

Area of Science:

• Genomics
• Cancer Biology
• Bioinformatics

Background:

• Bulk RNA sequencing obscures cancer cell-specific programs due to non-malignant cell admixture.
• Single-cell RNA sequencing (scRNA-seq) offers potential but faces challenges with data quality, annotation, and aggregation.
• Existing atlases often prioritize data volume over biological coherence, limiting interpretability.

Purpose of the Study:

• To develop a stringent integration framework for single-cell RNA sequencing (scRNA-seq) data, prioritizing representative malignant transcriptional states.
• To construct a high-quality, pan-cancer atlas of malignant cells for robust biological interpretation.
• To establish a scalable framework for tumor identity inference, cancer cell line benchmarking, and identification of genetic vulnerabilities.

Main Methods:

• Developed a stringent integration framework using Mahalanobis distance-based selection within batch-corrected latent space.
• Constructed a pan-cancer atlas of 135,424 high-quality malignant cells from 499 samples across 36 cancers.
• Utilized atlas-derived cancer signatures to assess tumor-cell line concordance and project ElasticNet models for gene dependency inference.

Main Results:

• Created a pan-cancer atlas prioritizing representative malignant transcriptional states from diverse adult and pediatric cancers.
• Demonstrated the utility of atlas-derived signatures for determining tumor-cell line concordance.
• Successfully projected models to infer cancer-specific gene dependencies, highlighting potential therapeutic targets.

Conclusions:

• The scTumor Atlas provides a scalable framework for accurate tumor identity inference.
• Enables robust benchmarking of cancer cell lines against primary tumors.
• Facilitates the systematic identification of genetic vulnerabilities across a wide range of cancers.