PolyAMiner-Bulk

PolyAMiner-Bulk decodes alternative polyadenylation (APA) dynamics from bulk RNA-seq data using an attention-based deep learning algorithm that recapitulates cleavage and polyadenylation site (C/PAS) sequence grammar, resolves overlapping C/PASs, and captures non-proximal-to-distal APA changes.

Key Features:

• De novo C/PAS detection: Identifies novel cleavage and polyadenylation sites without relying on predefined annotations.
• Overlapping C/PAS resolution: Deconvolutes overlapping cleavage and polyadenylation sites in bulk RNA-seq data.
• Non-proximal-to-distal APA capture: Detects APA changes beyond the most proximal and distal sites.
• Noisy 3' UTR coverage handling: Addresses inherent noise in 3' UTR coverage from bulk RNA-seq data.
• Attention-based deep learning: Employs an attention mechanism to accurately recapitulate C/PAS sequence grammar.
• Visualization generation: Generates visualizations to illustrate APA dynamics.

Scientific Applications:

• Tissue-specific APA analysis: Decodes novel tissue-specific alternative polyadenylation changes from bulk RNA-seq data.
• Disease-specific APA analysis: Identifies disease-specific alternative polyadenylation regulatory changes.
• Comparative APA profiling: Accurately identifies more APA changes compared with other bulk RNA-seq-based APA methods.

Methodology:

PolyAMiner-Bulk uses an attention-based deep learning algorithm to recapitulate C/PAS sequence grammar, resolve overlapping C/PASs, capture non-proximal-to-distal APA changes, and generate visualizations.