Multi-State Perfect Phylogeny Mixture Deconvolution and Applications to Cancer Sequencing
summary
Defines the multi-state perfect phylogeny mixture deconvolution problem as factoring a variant-allele frequency tensor into a shared usage matrix shared across slices and per-state perfect-phylogeny matrices, generalizing infinite-sites to an infinite-alleles model for SNVs+CNA. Proves hardness — NP-complete for k≥4 with two samples — and implements an enumerative solver that builds spanning trees labeled by valid descendant-set pairs and solves a constrained flow to fit confidence-intervaled frequencies; applies the pipeline to real bulk-sequencing data using phased copy-number calls.
AI-written summary · about