HapMuC

HapMuC detects somatic point mutations in impure and heterogeneous tumor samples by leveraging haplotype information from nearby heterozygous germline variants to improve mutation-calling accuracy.

Key Features:

• Utilization of Heterozygous Germline Variants: HapMuC incorporates haplotype phasing information from heterozygous germline variants near candidate mutations to distinguish true somatic changes from sequencing errors.
• Bayesian Hierarchical Framework: HapMuC employs a Bayesian hierarchical probabilistic model that constructs mutation and error generative models to prepare candidate haplotypes considering available heterozygous germline variants.
• Variational Bayesian Algorithm: HapMuC uses a variational Bayesian algorithm to infer haplotype frequencies and compute marginal likelihoods.
• Bayes Factor Scoring: HapMuC computes a Bayes factor to evaluate whether observed variations are true somatic mutations rather than sequencing artifacts.
• Read Realignment to Candidate Haplotypes: Observed sequencing reads are realigned to candidate haplotypes to reflect potential somatic changes.
• Validated Specificity and Sensitivity: Comparative analyses and simulations demonstrate superior specificity and sensitivity, validated on TCGA Mutation Calling Benchmark 4 and the COLO-829 cell line.

Scientific Applications:

• Low-frequency somatic mutation detection: Enables detection of low-frequency point mutations in impure and heterogeneous tumor samples.
• Tumor heterogeneity and evolutionary analysis: Supports studies of tumor heterogeneity and evolutionary dynamics by improving accuracy of mutation profiles.
• Method benchmarking and validation: Applicable to benchmarking somatic mutation callers and validating performance on datasets such as TCGA Mutation Calling Benchmark 4 and COLO-829.

Methodology:

Candidate haplotypes are prepared by considering available heterozygous germline variants; observed sequencing reads are realigned to these candidate haplotypes; haplotype frequencies and marginal likelihoods are inferred using a variational Bayesian approach; and a Bayes factor is derived to evaluate the likelihood of true somatic mutations.