mobster-R

mobster-R performs tumour subclonal deconvolution from bulk DNA whole-genome sequencing data by modeling the site frequency spectrum of somatic mutations to characterize subclonal composition and selection.

Key Features:

• Model-based approach: Uses the site frequency spectrum of somatic mutations with an explicit neutral-evolution model to deconvolute subclones and infer clonal expansions from bulk WGS data.
• Integration of machine learning and population genetics: Combines principles from machine learning with population-genetics modeling to represent neutral and positive selection processes simultaneously.
• Comprehensive analysis functions: Provides functions for plotting sequencing data, fitting models, assessing confidence of model fits, and performing evolutionary analyses of subclonal structures.
• Empirical validation: Has been applied to simulated datasets and real-world cancer whole-genome sequencing data, including breast cancer samples.

Scientific Applications:

• Tumour evolutionary reconstruction: Infers subclonal architecture and clonal expansions from bulk WGS to study tumour evolution.
• Detection of selection: Differentiates neutral evolutionary processes from positive selection within tumour populations using population-genetics-informed models.
• Quantification of intra-tumour heterogeneity: Characterizes genetic diversity and subclonal composition from bulk somatic mutation data.
• Translational research support: Provides subclonal decomposition results that can inform studies of disease progression and personalized treatment strategies.

Methodology:

Model-based analysis of the site frequency spectrum of somatic mutations with an explicit neutral-evolution component combined with machine-learning and population-genetics modeling, including model fitting, confidence assessment, and plotting.