Motto

Motto converts position weight matrices (PWMs) into compact wildcard-style consensus sequences that minimize information loss for motif representation and interpretation.

Key Features:

• Mathematical framework: Utilizes mutual information theory and Jensen-Shannon divergence to formalize conversion from PWMs to consensus sequences.
• Sequence Motto representation: Produces "sequence Motto" wildcard-style consensus sequences for motifs derived from nucleotides, amino acids, and customized characters.
• Information-preserving conversion: Minimizes information loss when transforming PWMs into compact consensus sequences for interpretation and motif searching.
• Alphabet support: Handles motifs from nucleotide alphabets, amino acid alphabets, and user-defined custom characters.
• Binding site identification: Demonstrated effectiveness in identifying transcription factor binding sites across the human genome.
• Benchmarking: Evaluated against PWM scanning by FIMO using area under the precision-recall curve (AUPRC) and statistical testing.
• Comparative performance: Achieved a mean AUPRC of 0.81 and significantly outperformed maximal positional weight, Cavener's method, and minimal mean square error (p < 0.01).

Scientific Applications:

• Transcription factor binding site identification: Identifying TF binding sites in genomic sequences, demonstrated for 1,156 human transcription factors.
• Motif interpretation and searching: Generating concise consensus sequences for interpreting motif information and for searching motif matches.
• Method benchmarking and comparison: Evaluating and comparing motif representation methods using AUPRC and statistical significance tests against FIMO and alternative methods.
• Sequence analysis in genomics: Producing compact motif summaries for downstream sequence analysis tasks in genomics and bioinformatics.

Methodology:

Conversion of PWMs to consensus sequences using a framework based on mutual information theory and Jensen-Shannon divergence; benchmarking via AUPRC against PWM scanning by FIMO and comparison to maximal positional weight, Cavener's method, and minimal mean square error on 1,156 human TFs.