SiCloneFit

SiCloneFit reconstructs clonal populations, genotypes, and phylogenies from single-cell DNA sequencing (SCS) data using a nonparametric Bayesian framework to model intra-tumor heterogeneity.

Key Features:

• Nonparametric Bayesian joint inference: Clusters single cells into distinct clones and jointly infers clone genotypes and clonal phylogeny using a nonparametric Bayesian model.
• Tree-structured Chinese Restaurant Process (CRP): Uses a tree-structured CRP as a prior to define the number and composition of clonal populations.
• Clonal phylogeny with finite-site evolutionary model: Models clonal evolution with a clonal phylogeny combined with a finite-site model that accounts for mutation recurrence and losses.
• Probabilistic error modeling: Accounts for false positives and false negatives, including allelic dropout, in SCS data through probabilistic error handling.
• Cell doublet modeling: Models cell doublets explicitly using a Beta-binomial distribution.
• Advanced sampling algorithm: Explores the joint posterior with a Gibbs sampling algorithm incorporating partial reversible-jump and partial Metropolis–Hastings updates.
• Posterior support measures: Provides measures of support for inferred clones, genotypes, and phylogenies derived from the posterior distribution.

Scientific Applications:

• Cancer clonal evolution analysis: Reconstruction of clonal populations and their evolutionary relationships in tumor samples from SCS data.
• Intratumor heterogeneity characterization: Quantification and interpretation of intra-tumor heterogeneity by resolving clone genotypes and frequencies.
• Method validation on experimental and synthetic data: Assessment of inference robustness and support measures using both experimental and simulated SCS datasets.
• Support for diagnostic and therapeutic research: Generation of clonal and phylogenetic hypotheses that can inform studies of tumor progression and treatment resistance.

Methodology:

Performs nonparametric Bayesian inference with a tree-structured CRP prior and a finite-site model of evolution, probabilistically models FP/FN errors and allelic dropout and models doublets with a Beta-binomial distribution, and samples the joint posterior using a Gibbs sampler with partial reversible-jump and partial Metropolis–Hastings updates to jointly infer clones, genotypes, and phylogeny.