SCARLET

SCARLET reconstructs tumor phylogenies from single-cell DNA sequencing data by explicitly modeling copy-number–supported mutation losses to improve phylogenetic inference in CNA-rich tumors.

Key Features:

• Loss-supported evolutionary model: Implements a model that constrains mutation losses to loci with evidence of decreased copy number (CNAs).
• Model generalization: Generalizes traditional infinite sites and Dollo models to permit losses only when supported by copy-number changes.
• SNV-based phylogenetic markers: Uses single-nucleotide variants (SNVs) as markers for phylogenetic reconstruction while modeling SNV deletion by CNAs.
• Probabilistic error model: Employs a probabilistic model to account for sequencing errors and allele dropout.
• Simulation-based validation: Demonstrated improved phylogenetic accuracy and correction of SNV errors on simulated single-cell sequencing data.
• Empirical application: Applied to single-cell sequencing from a metastatic colorectal cancer patient to produce phylogenies consistent with copy-number data and infer monooclonal metastasis seeding.

Scientific Applications:

• Tumor phylogeny reconstruction: Reconstructs clonal trees from single-cell DNA sequencing data in the presence of copy-number aberrations (CNAs).
• Loss-event interpretation: Distinguishes mutation loss events that are supported by overlapping CNAs from unsupported losses.
• Metastasis seeding analysis: Infers tumor seeding patterns and clonal dynamics in metastatic samples, exemplified by metastasis seeding in colorectal cancer.
• SNV genotype correction: Improves correction of SNV calls in single-cell datasets by incorporating CNA-aware loss modeling and error/allele-dropout modeling.

Methodology:

Combines a novel loss-supported evolutionary model (generalizing infinite sites and Dollo) that constrains mutation losses to loci with decreased copy number, uses SNVs as phylogenetic markers, and incorporates a probabilistic model for sequencing errors and allele dropout.