CancerCellNet

CancerCellNet assesses and classifies cancer models—including cell lines, patient-derived xenografts (PDX), genetically engineered mouse models (GEMMs), and tumoroids—by quantifying their transcriptional fidelity to native tumors across 22 tumor types and 36 subtypes using platform-agnostic transcriptional data.

Key Features:

• Implementation: Provided as an R package for computational analysis.
• Algorithms: Employs machine learning algorithms to classify and score transcriptional similarity.
• Platform-agnostic data handling: Accommodates bulk RNA sequencing and microarray data across different species.
• Model scope: Evaluates multiple cancer model types, including cell lines, PDXs, GEMMs, and tumoroids.
• Tumor coverage: Compares models against reference profiles for 22 tumor types and 36 subtypes.
• Applied dataset scale: Analysis reported on 657 cancer cell lines, 415 PDXs, 26 GEMMs, and 131 tumoroids.
• Validation: Includes rigorous validation with independent datasets and experimental confirmation via immunofluorescence.
• Output capabilities: Identifies models with high transcriptional fidelity, detects discrepancies between annotations and classifications, and highlights gaps in model representation.

Scientific Applications:

• Model fidelity assessment: Measures how closely cancer models mimic native tumor transcriptional profiles.
• Model selection: Guides selection of models for biological studies and translational research based on transcriptional similarity.
• Comparative evaluation: Compares transcriptional fidelity across model types (cell lines, PDXs, GEMMs, tumoroids) and tumor types/subtypes.
• Gap identification: Identifies underrepresented tumor models to inform future model development and research priorities.
• Annotation refinement: Detects mismatches between model annotations and molecular classifications to refine modeling strategies.

Methodology:

Implemented as an R package that applies machine learning algorithms to platform-agnostic transcriptional data (bulk RNA-seq and microarrays) across species to evaluate similarity between cancer models and reference profiles for 22 tumor types and 36 subtypes.