Model-based subclonal deconvolution from bulk sequencing.
MOBSTER is a package that implements a statistical model for tumour subclonal deconvolution which combines Dirichlet mixtures with distributions predicted by theoretical population genetics. The model implemented is a parametric (finite) mixture with multiple Beta components, and one Pareto power law tail. Beta components capture mutations that rise up in frequency by positive selection, and the Pareto tail models alleles resulting from neutral, intra-clone, evolutionary dynamics.
