Strelka

Strelka detects somatic single nucleotide variants (SNVs) and small insertions and deletions (indels) in matched tumor-normal whole-genome and exome sequencing data to enable accurate somatic variant identification in cancer genomics.

Key Features:

• Somatic variant detection: Calls somatic SNVs and small indels from matched tumor-normal whole-genome (WGS) and whole-exome (WES) sequencing data.
• Bayesian allele-frequency model: Implements a Bayesian approach that represents continuous allele frequencies for both tumor and normal samples.
• Normal mixture model: Models the normal sample as a mixture of germline variation and noise.
• Tumor composition model: Represents the tumor sample as the normal sample genotype plus additional somatic variations.
• Purity robustness: Maintains high sensitivity at elevated levels of tumor impurity without requiring prior tumor purity estimates.
• Genotype-aware modeling: Incorporates the expected genotype configuration of normal samples to distinguish somatic events.
• Improved accuracy: Achieves higher sensitivity and accuracy compared with diploid genotype likelihoods and general allele-frequency tests.

Scientific Applications:

• Cancer somatic variant discovery: Precise identification of somatic SNVs and indels for cancer genomics studies.
• Tumor heterogeneity and evolution: Detection of low-frequency somatic variants in impure and heterogeneous tumor samples to study tumor evolution.
• Therapeutic target identification: Identification of candidate somatic variants that can inform potential therapeutic targets and personalized treatment strategies.
• WGS/WES analyses: Application to matched tumor-normal WGS and WES datasets for comprehensive somatic variant calling.

Methodology:

Uses a Bayesian model representing continuous allele frequencies for tumor and normal, models the normal sample as a mixture of germline variation and noise and the tumor as the normal genotype plus somatic variants, incorporates expected normal genotype configurations, and does not require prior tumor purity estimates.