ISOWN

ISOWN identifies somatic single nucleotide variants (SNVs) in cancer genomes without requiring matched normal tissue controls by applying supervised machine learning to next-generation sequencing data.

Key Features:

• Somatic SNV calling without matched normal: Detects somatic single nucleotide variants (SNVs) in tumor samples when matched normal tissue is unavailable.
• Supervised machine learning: Uses supervised learning classifiers to distinguish somatic mutations from germline polymorphisms.
• Next-generation sequencing input: Operates on NGS data, including deep targeted sequencing and whole-exome sequencing datasets.
• Cross-tissue validation: Evaluated across six different cancer types using approximately 1,600 samples.
• Sample type compatibility: Assessed on cell lines, fresh frozen tissues, and formalin-fixed paraffin-embedded (FFPE) tissues.
• Performance metrics: Reported correct classification rates of 95%–98% and F1-measure ranges of 75.9%–98.6% depending on tumor type.

Scientific Applications:

• Cancer genomics: Identification of somatic mutations in tumor-only sequencing studies to profile cancer genomes.
• Tumor-type comparative studies: Cross-cancer assessments and comparative analyses across multiple tumor types.
• FFPE and archived sample analysis: Somatic variant calling in FFPE and other archived specimens where matched normals are often absent.
• Preliminary variant discovery for personalized medicine: Supporting somatic variant discovery in contexts that inform cancer genetics and precision oncology.

Methodology:

Supervised machine learning classification applied to next-generation sequencing data (deep targeted and whole-exome), with validation on ~1,600 samples across six cancer types including cell lines, fresh frozen, and FFPE tissues.