COOBoostR

COOBoostR predicts the tissue or cell-of-origin of cancer samples by applying an extreme gradient boosting algorithm to regional somatic mutation density and chromatin mark features.

Key Features:

• Data Integration: Integrates ChIP-seq-based chromatin data with regional somatic mutation density from normal cells/tissue, precancerous lesions, and cancer types.
• Machine Learning Algorithm: Uses an extreme gradient boosting algorithm to model relationships between chromatin features and somatic mutation density.
• Chromatin Mark Ranking: Ranks chromatin marks from diverse tissue and cell types by their explanatory power for the somatic mutation density landscape of a sample.
• Performance Metrics: Demonstrates faster prediction speed than random forest-based methods while reporting accuracies of 76.99% for normal cells/tissue, 95.65% for precancerous lesions, and 89.39% for cancer cells.
• Dynamic Mutation Accumulation: Indicates a dynamic process of somatic mutation accumulation at the normal tissue or cell stage linked with changes in open chromatin marks and enhancer sites.

Scientific Applications:

• Tissue/Cell-of-Origin Inference: Predicts the likely tissue or cell origin of tumor samples using somatic mutation and chromatin data.
• Study of Early Tumorigenesis: Enables investigation into mutation accumulation and chromatin dynamics during normal, precancerous, and cancer stages.
• Support for Translational Research: Provides information relevant to understanding mechanisms that may inform targeted therapy research and diagnostic precision.

Methodology:

Integrates ChIP-seq-based chromatin data with regional somatic mutation density, applies extreme gradient boosting to model explanatory relationships, ranks chromatin marks by their ability to explain mutation density, and compares predictive speed and accuracy against random forest-based methods.