Meltos: multi-sample tumor phylogeny reconstruction for structural variants.

Area of Science:

• Computational biology
• Genomics
• Cancer research

Background:

• Constructing accurate tumor phylogeny trees is crucial for understanding cancer evolution.
• Somatic structural variants (SVs) provide valuable insights but are challenging to integrate into phylogenetic analyses.
• Existing methods often struggle to reliably incorporate SV data into tumor lineage reconstruction.

Purpose of the Study:

• To introduce Meltos, a novel computational framework for building tumor phylogeny trees using somatic structural variants (SVs) across multiple samples.
• To leverage existing somatic single nucleotide variant (SNV) SNV-based phylogeny to guide the identification and placement of SVs.
• To develop a robust method for estimating variant allele fractions (VAFs) of SV events using multiple genomic read signals.

Main Methods:

• Meltos utilizes a tumor phylogeny tree built from SNVs as a guide for SV calling and assignment.
• It employs a novel optimization formulation to produce a comprehensive tumor lineage tree.
• The framework incorporates multiple genomic read signals for SV breakpoint detection and uses probabilistic methods for VAF estimation.

Main Results:

• Meltos was evaluated on simulated datasets and two real cancer datasets (liposarcoma and breast cancer).
• The framework successfully refined SNV-based trees by incorporating SV information, placing high-confidence SVs accurately.
• Meltos demonstrated flexibility in VAF estimation, either directly from genomic data or using copy number corrected estimates.

Conclusions:

• Meltos provides a powerful new approach for constructing more comprehensive tumor phylogeny trees by integrating SV data.
• The framework enhances the accuracy and resolution of cancer evolutionary models.
• Meltos facilitates a deeper understanding of tumor heterogeneity and progression.