Meltos

Meltos reconstructs and refines tumor phylogenies by assigning somatic structural variants (SVs) to branches of somatic single nucleotide variant (SNV)-based lineage trees using probabilistic variant allele frequency (VAF) estimation from whole genome sequencing data.

Key Features:

• Integration of SNV and SV Data: Uses SNV-based phylogenetic trees to guide placement and assignment of somatic structural variants (SVs) across a tumor lineage while accounting for potential differences in evolutionary trajectories between SNVs and SVs.
• Probabilistic Framework for VAF Estimation: Employs a probabilistic model to estimate variant allele frequencies (VAFs) of SV events from whole genome sequencing data by jointly assessing multiple genomic read signals at SV breakpoints.
• Combinatorial Algorithm for Tree Refinement: Implements a combinatorial algorithm grounded in maximum parsimony to assign clusters of SVs onto branches of an existing lineage tree, permitting modest adjustments to tree topology to accommodate SV data.
• Flexibility in Data Utilization: Supports either direct VAF estimation from genomic data or the use of copy number corrected VAF estimates.

Scientific Applications:

• Simulated datasets: Validated on simulations involving five genomes to demonstrate refinement of SNV-based trees with SV information.
• Liposarcoma tumor dataset: Applied to a liposarcoma tumor dataset containing validated structural variation events.
• Multi-sample breast cancer data: Applied to multi-sample breast cancer data (Yates et al., 2015) combining validated SV calls with deep targeted sequencing of somatic SNVs.

Methodology:

Uses an SNV-only tree as a basis, applies probabilistic VAF estimation by integrating multiple genomic read signals at SV breakpoints, and formulates a combinatorial optimization under a maximum parsimony principle to assign SV clusters and refine tree topology.