Inferring models of multiscale copy number evolution for single-tumor phylogenetics

Area of Science:

• Computational Biology
• Cancer Genomics
• Evolutionary Medicine

Background:

• Phylogenetic algorithms are increasingly used in cancer research to model tumor evolution.
• Accurate tumor phylogenies require quantitative models that incorporate complex genetic mechanisms like chromosome abnormalities and patient-specific heterogeneity.
• Existing methods often oversimplify tumor evolution by assuming uniform rates of genomic gain and loss.

Purpose of the Study:

• To develop a novel framework for inferring tumor progression models from single-cell gene copy number data.
• To introduce algorithms capable of handling variable rates of genomic gain and loss, including chromosome abnormalities and genome duplications.
• To concurrently estimate mutation-specific and tumor-specific event rates alongside phylogenetic tree reconstruction.

Main Methods:

• Proposed a framework for inferring tumor progression from single-cell gene copy number data.
• Developed a new algorithm for identifying parsimonious combinations of single gene and chromosome events, extended with dynamic programming for genome duplications.
• Implemented an expectation-maximization (EM)-like method for concurrent estimation of event rates and tree reconstruction.

Main Results:

• Applied the algorithms to cervical cancer data, identifying key genomic events in disease progression.
• Achieved improved prediction accuracy in classification experiments on cervical and tongue cancer datasets.
• Demonstrated enhanced prediction for metastasis of primary cervical cancers and survival for tongue cancer.

Conclusions:

• The developed framework and algorithms provide a more realistic and accurate approach to modeling tumor evolution.
• The method successfully integrates complex genetic events and variable rates, improving phylogenetic inference from copy number data.
• The findings have implications for understanding cancer progression and improving diagnostic and prognostic predictions.