kmer-SVM

kmer-SVM applies support vector machine classification to k-mer sequence features to predict transcription factor binding sites and regulatory elements from genomic datasets generated by massively parallel sequencing technologies such as ChIP-seq and DNase-seq.

Key Features:

• Support Vector Machine Integration: Employs a support vector machine (SVM) for sequence classification and predictive modeling of genomic regions.
• k-mer Sequence Features: Uses k-mer representations of short, contiguous nucleotide sequences as primary input features for the SVM.
• Predictive k-mer Combinations: Identifies combinations of short transcription factor–binding k-mers that are predictive of tissue-specific genomic assay outcomes.
• Compatibility with ChIP-seq and DNase-seq: Analyzes sequences derived from ChIP-seq and DNase-seq assays to associate sequence patterns with binding and open chromatin.
• Recovery of Known Binding Sites: Recovers previously known transcription factor binding sites from genomic datasets.
• Discovery of Novel Sequence Features: Reveals novel sequence features amenable to further experimental validation.
• Identification of Accessory and Repressive Elements: Detects sequence patterns corresponding to accessory factors and repressive regulatory elements.

Scientific Applications:

• Enhancer Prediction: Discriminatively predicts mammalian enhancers from DNA sequence.
• Validation of Known Binding Sites: Validates genomic experiments by recovering known transcription factor binding sites.
• Discovery of Novel Sequence Features: Supports discovery of novel sequence motifs and features for experimental follow-up.
• Identification of Accessory Factors and Repressive Elements: Identifies accessory factors and repressive sequence elements involved in gene regulation.

Methodology:

Uses k-mer sequence features as input to a support vector machine to classify sequences and identify predictive k-mer combinations corresponding to transcription factor binding sites and regulatory elements.