scPolyA enables efficient, interactive and agent-guided alternative polyadenylation analysis in single-cell and spatial transcriptomes

scPolyA maps alternative polyadenylation at single-cell and spatial RNA 3'-ends, achieving higher reference-atlas agreement than most de novo methods across diverse protocols and platforms.

• Why it matters: APA adds regulatory variation not captured by gene-level expression, and existing 3'-end analyses struggle with sparse poly(A) evidence and limited interactive, question-driven workflows.
• What they did: The framework unifies de novo poly(A)-site discovery with cell/spot-resolved terminal-site usage, 3' UTR length, and intronic polyadenylation, using DeepCPS for nucleotide-resolution cleavage prediction and an AI agent for natural-language analysis execution.
• The result: In cutaneous squamous cell carcinoma, scPolyA detects recurrent 3' UTR shortening in tumour keratinocytes across all ten patients and identifies tissue-level contrasts shaped by cellular composition.

AI summary of the abstract · may contain errors · not medical advice · read the paper.