CTDPathSim

CTDPathSim computes similarity scores between patient tumor samples and cancer cell lines using a pathway activity–based approach that integrates deconvoluted gene expression and DNA methylation profiles and is implemented as an R package.

Key Features:

• Pathway activity–based similarity scoring: Computes similarity between tumor samples and cancer cell lines using pathway-level activity profiles.
• Multi-omics integration: Integrates sample-specific gene expression and DNA methylation data for similarity assessment.
• Deconvolution of bulk tumors: Applies a deconvolution method to derive cell type–specific DNA methylation and gene expression profiles accounting for tumor heterogeneity.
• Deconvoluted sample profiles: Generates deconvoluted methylation and expression profiles for tumor samples to enable more precise comparisons to cell lines.
• Validation on public resources: Validated using breast and ovarian cancer data from The Cancer Genome Atlas (TCGA) and cancer cell line data from the Cancer Cell Line Encyclopedia (CCLE).
• Drug-response association: Associates high sample–cell line similarity with comparable drug responses, including Paclitaxel, Vinorelbine, and Mitomycin-c.
• Performance against correlation methods: Outperforms genome-wide correlation–based methods in recapitulating known drug responses and identifying significant cell lines within the same cancer types.
• Clinical biomarker capability: Identifies aligned cell lines that serve as clinical biomarkers for patient survival.

Scientific Applications:

• Cell line selection for preclinical testing: Selects cell lines that better represent patient tumor pathway activity for use in preclinical drug evaluation.
• Prediction of drug response: Infers comparable drug responses between patient samples and cell lines for FDA-approved drugs such as Paclitaxel, Vinorelbine, and Mitomycin-c.
• Patient stratification and prognostic biomarkers: Uses aligned cell lines as clinical biomarkers to stratify patients and relate to survival outcomes.
• Benchmarking similarity methods: Provides a pathway-based alternative to genome-wide correlation approaches for comparing tumors and cell lines.
• Drug repurposing support: Aids identification of new uses for existing drugs by linking tumor samples to cell lines with known drug sensitivities.

Methodology:

Apply deconvolution to derive cell type–specific DNA methylation and gene expression profiles, compute deconvoluted methylation and expression profiles for tumor samples, and calculate pathway activity–based similarity scores between tumor samples and cancer cell lines; validation performed using TCGA (breast and ovarian) and CCLE data.