discrover

discrover identifies discriminative sequence motifs in nucleic acid sequences using hidden Markov models and mutual information to distinguish motifs between positive and negative example sets.

Key Features:

• Discriminative learning with HMMs: Employs a discriminative learning approach based on hidden Markov models to identify sequence patterns that differ between datasets.
• Mutual information objective: Uses mutual information as the objective function to quantify the association between condition and motif occurrence.
• Performance and efficiency: Demonstrated higher accuracy and faster processing speeds in systematic comparisons with published motif-finding tools.
• Case studies across technologies: Applied to data types including ChIP-Seq, RIP-Chip, and PAR-CLIP, covering transcription factors in embryonic stem cells and RNA-binding proteins.
• Complex data configurations: Handles binary contrasts and more complex experimental configurations for genome- and transcriptome-scale analyses.
• Integration of repeat experiments: Makes use of available repeat experiments to enhance robustness and reliability of motif discovery.

Scientific Applications:

• Transcription factor motif discovery: Identification of transcription factor motifs from ChIP-Seq data, including studies in embryonic stem cells.
• RNA-binding protein motif discovery: Discovery of RNA-binding protein motifs from RIP-Chip and PAR-CLIP datasets.
• Alternative splicing factor analysis: Analysis of alternative splicing factors such as RBM10 to identify motifs relevant to splicing regulation.
• Genome- and transcriptome-scale motif analysis: Discriminative motif analysis across genome- and transcriptome-scale datasets and complex experimental designs.
• Robust motif identification with replicates: Use of repeat experiments to improve the robustness of discovered motifs.

Methodology:

Discriminative learning using hidden Markov models with mutual information as the objective function; systematic comparisons with published motif-finding tools; support for binary and multi-condition configurations and integration of repeat experiments.