Abstract

Cancer cells constantly evolve accumulating somatic mutations. To describe the tumor evolution process, we develop the Tumor Evolution Decoder (TED), a novel algorithm for constructing phylogenetic tree based on somatic mutation profiles of tumor subclones or single cells. TED takes a unique strategy that reduces the total number of duplicated mutations and dropout mutations in the tumor evolution process, which has not been explored by previous phylogenetic tree methods. TED allows multiple types of somatic mutations as input, such as point mutations, copy number alterations, gene fusion, and their combinations. Theoretical properties of TED are derived while its numerical performance is examined using simulated data. We applied TED to analyze single-cell sequencing data from an essential thrombocythemia tumor and a clear cell renal cell carcinoma, to investigate the ancestral relationships between tumor cells, and found genes related to disease initialization and development mutated in the early steps of evolution. We also applied TED to the subclones of a breast invasive carcinoma and provided important insights on the evolution and metastasis of the tumor.