DrCancer

DrCancer predicts cancer-causing single nucleotide polymorphisms (SNPs), focusing on missense (non-synonymous) variants to distinguish pathogenic from benign mutations.

Key Features:

• Machine Learning Approach: Employs a Support Vector Machine (SVM) classifier trained on 3,163 known cancer-causing variants and an equal number of neutral polymorphisms.
• Analysis Scope: Analyzes non-synonymous (missense) SNPs that result in amino acid changes within protein-coding regions.
• Data Sources: Integrates high-throughput genotyping and sequencing data for model training and prediction.
• Predictive Performance: Achieves 93% accuracy, a correlation coefficient of 0.86, and an area under the ROC curve (AUC) of 0.98.
• Comparative Superiority: Outperforms existing algorithms SIFT and CHASM in prediction accuracy and correlation coefficient.

Scientific Applications:

• Cancer Research: Helps identify missense variants that may contribute to oncogenesis and supports studies of genetic drivers of cancer.
• Genetic Screening: Supports clinical genetic screening for early diagnosis, risk assessment, and personalized treatment planning by identifying potential cancer-causing SNPs.

Methodology:

Uses a machine learning framework that integrates large-scale genetic data to train an SVM classifier and analyzes non-synonymous SNPs to assess their functional impact on proteins involved in cell proliferation and cancer development.