SNVMix

SNVMix detects single nucleotide variants (SNVs) from next-generation sequencing (NGS) data derived from tumor tissues using probabilistic binomial mixture models that account for altered ploidy and variable tumor cellularity.

Key Features:

• Probabilistic Binomial Mixture Model: Employs a binomial mixture model to infer SNVs from NGS data and accommodate deviations in allelic frequencies due to genomic instability.
• Allelic Count Modeling (EM): Models allelic counts directly and uses an expectation maximization (EM) algorithm to infer SNVs and model parameters.
• Quality-Weighted Nucleotide Inference: Incorporates nucleotide and mapping qualities to probabilistically weight each read's contribution to SNV inference.
• Combined Filtering and Weighting: Provides an implementation that filters low-quality data while applying probabilistic weighting of nucleotide qualities to refine detection.
• Tumor-Specific Considerations: Accounts for altered ploidy and variable tumor cellularity common in tumor-derived sequencing data.

Scientific Applications:

• Somatic SNV detection in cancer genomics: Detects single nucleotide variants in tumor-derived NGS datasets, including samples with genomic instability and variable cellularity.
• Benchmarking and validation: Evaluated on 16 ovarian cancer RNASeq datasets with matched genotyping arrays and on a human breast cancer genome sequenced to over 40x coverage to assess accuracy.

Methodology:

Millions of short sequence reads are aligned to a reference genome, after which probabilistic binomial mixture models (including an EM-based allelic count model) are applied, integrating nucleotide and mapping quality metrics via probabilistic weighting and optional low-quality data filtering.