MFCNV

MFCNV detects copy number variations (CNVs) in tumor genomes from next-generation sequencing (NGS) data to improve the accuracy of CNV identification for cancer research.

Key Features:

• Intrinsic correlation modeling: Accounts for intrinsic correlations among adjacent genomic positions to improve CNV signal reliability.
• Multi-feature per-bin evaluation: Calculates read depth, GC-content bias, base quality, and correlation value for each genome bin to inform CNV calling.
• Neural network algorithm: Uses a neural network to model the joint effects of multiple influencing factors on CNV detection.
• Performance metrics: Reports high sensitivity, precision, and F1-score for CNV detection relative to existing methods.

Scientific Applications:

• Cancer genomics: Detects CNVs in tumor genomes to support studies of tumor development and genomic drivers of cancer.
• Therapeutic target discovery: Identifies CNVs that may inform potential therapeutic targets and biomarkers in oncology research.
• Single-cell sequencing analysis: Can be applied to single-cell sequencing data to broaden CNV analysis at the single-cell level.
• Comprehensive CNV discovery: Reveals CNVs that may be missed by other methods to provide a more complete view of genomic alterations in cancer cells.

Methodology:

MFCNV computes per-bin features (read depth, GC-content bias, base quality, correlation value), models intrinsic adjacent-position correlations, and applies a neural network trained on simulated and real datasets for CNV detection; program files include scripts for training the neural network, detecting CNVs in simulated and real datasets, extracting eigenvalues, and transforming text to MATLAB (MAT) format, while test data include MAT-format real and simulation data, original BAM files, and reference sequences.