Deep Sequence and Shape Motif (DESSO)

DESSO identifies cis-regulatory sequence and DNA shape motifs and predicts transcription factor binding sites to elucidate TF-DNA binding and tethering interactions in the human genome.

Key Features:

• Deep Learning Integration: DESSO employs deep neural networks and surpasses DeepBind in predicting cis-regulatory motifs.
• Binomial Distribution Model: DESSO integrates a binomial distribution model with neural networks to predict transcription factor binding sites.
• Comprehensive Dataset Analysis: The method was tested on 690 human ENCODE ChIP-sequencing datasets for motif prediction benchmarking.
• Discovery of Tethering Interactions: DESSO facilitates identification of protein-protein-DNA tethering interactions and identified 61 putative tethering interactions among 100 expressed TFs in the K562 cell line.
• DNA Shape Feature Integration: DESSO incorporates DNA shape features, improving prediction of TF-DNA binding and revealing potential shape motifs associated with human transcription factors.

Scientific Applications:

• Transcription factor binding site mapping: Predicts sequence-based and shape-based motifs from ChIP-sequencing data to map TF binding sites.
• Elucidation of TF-DNA recognition rules: Reveals sequence and DNA shape motifs that inform mechanisms of transcriptional regulation.
• Identification of tethering interactions: Detects protein-protein-DNA tethering interactions among transcription factors, as demonstrated in the K562 cell line.
• Regulatory and disease genomics: Provides motif-level insights relevant to studies of gene regulation and disease-associated regulatory variants.

Methodology:

DESSO applies deep neural networks combined with a binomial distribution model to analyze sequence data and incorporates DNA shape information for motif and transcription factor binding prediction.