SAMSVM

SAMSVM detects and filters misaligned reads in Sequence Alignment/Map (SAM)-formatted sequences to reduce false positives in variant calling and improve alignment-based genomic analyses.

Key Features:

• Misalignment Detection: Identifies and filters misaligned reads from SAM-formatted sequences to minimize false positives during variant calling.
• Vector Space Representation: Represents multiple features of SAM-formatted sequences as vectors in a multi-dimensional space for classification.
• LIBSVM Integration: Implements support vector classification using LIBSVM (Chang and Lin) to classify aligned reads.
• Validation and Accuracy: Validated by cross-validation on two simulated datasets, achieving accuracies of 0.89–0.97 and F-scores of 0.77–0.94 across 14 groups with mutation rates 0.001–0.1.
• Practical Application: Applied to real sequencing data to filter misaligned reads and thereby improve the accuracy of downstream variant calling.

Scientific Applications:

• Variant calling: Reduces false positives in variant calling by removing misaligned reads from SAM files.
• Mutation detection and genetic variation studies: Supports mutation detection and genetic variation analyses by improving alignment quality used in downstream interpretation.
• Massive parallel sequencing data processing: Processes both simulated and real data generated by massive parallel sequencing technologies for alignment-quality control.

Methodology:

Represents SAM features as vectors in multi-dimensional space and applies support vector machine classification via LIBSVM, with performance assessed by cross-validation on two simulated datasets.