SATORI

SATORI models and infers cooperativity between regulatory features in genomic sequences using convolutional neural network layers combined with a self-attention mechanism to detect transcription factor (TF) interactions relevant to gene regulation.

Key Features:

• Self-Attention Mechanism: Integrates a self-attention mechanism with convolutional layers to capture long-range dependencies and provide a global view of relevant sequence regions without requiring intensive post-processing.
• Feature Interaction Detection: Detects statistically significant TF-TF interactions and identifies cooperativity between regulatory elements directly from sequence data.
• Versatility Beyond TF-TF Interactions: Uses an attention-based architecture that can be applied to detect other types of feature interactions where similar attention mechanisms are applicable.

Scientific Applications:

• TF-TF interaction discovery: Identification of statistically significant transcription factor–transcription factor interactions from genomic sequences.
• Regulatory network elucidation: Revealing cooperative interactions among regulatory elements to inform models of gene expression control.
• Large-scale genomic analysis: Extraction of interaction patterns from large genomic datasets with minimal post-processing.

Methodology:

Combines convolutional neural network layers with a self-attention mechanism to extract features from genomic sequences and infer regulatory feature cooperativity.