DeepYY1

DeepYY1 predicts YY1-mediated chromatin loops using deep learning to identify sequence features that underlie enhancer–promoter interactions and YY1 dimerization.

Key Features:

• Identification of YY1-mediated loops: Predicts whether pairs of YY1 motifs form chromatin loops mediated by Yin Yang 1 (YY1).
• Deep learning model: Uses deep learning techniques to learn predictive sequence patterns associated with loop formation.
• Sequence embedding: Employs the word2vec algorithm to encode sequence and motif features for model input.
• Motif-pair prediction: Specifically predicts loop formation between pairs of YY1 motifs.
• Performance: Achieves area under the curve (AUC) ≥ 0.93 on both training and testing datasets across multiple cell types.
• Sequence significance: Highlights that specific sequence information is critical for YY1-mediated loop formation.
• Biological insight: Analyzes the distribution of replication origin sites within predicted YY1-mediated loops.
• Cross-cell-type evaluation: Trained and evaluated on datasets from different cell types.

Scientific Applications:

• Mapping enhancer–promoter interactions: Identifies YY1-mediated enhancer–promoter loops in genomic sequence data.
• Gene regulation studies: Supports investigation of gene regulation mechanisms involving YY1 dimerization.
• Chromatin architecture analysis: Aids studies of chromatin looping and 3D genome organization across cell types.
• Replication origin research: Enables examination of replication origin site distribution within chromatin loops and implications for genomic stability.
• Comparative cell-type analysis: Facilitates comparison of YY1-mediated interactions across different cell types.

Methodology:

Implements deep learning models using word2vec-based sequence embeddings to predict whether pairs of YY1 motifs form chromatin loops, with performance evaluated by AUC (≥0.93) on training and testing datasets across multiple cell types and an emphasis on sequence information.