PolyA-miner

PolyA-miner analyzes differential alternative polyadenylation (APA) from 3' sequencing (3'Seq) data to detect and quantify gene-level 3' untranslated region (3'UTR) changes, including novel polyadenylation sites.

Key Features:

• APA quantification: Accurately assesses differential alternative polyadenylation using 3'Seq data.
• APA matrix abstraction: Abstracts genes into APA matrices for downstream analysis.
• Iterative consensus NMF clustering: Uses iterative consensus non-negative matrix factorization (NMF) based clustering to infer differential APA usage.
• Vector projection analysis: Applies vector projections to capture proximal-to-distal, non-proximal-to-non-distal, and other APA switches at the gene level.
• Novel site discovery: Identifies novel APA sites not present in reference annotations.
• Protocol and dataset compatibility: Validated on PolyA-seq, PAC-seq, and other 3'Seq protocols including MAQC brain, UHR PolyA-seq, glioblastoma NUDT21 knockdown PAC-seq, mouse hippocampal, and human stem cell-derived neuron PAC-seq datasets.
• Enhanced sensitivity: Detected more than twice the number of genes with APA changes in glioblastoma NUDT21 knockdown data compared to initial reports.

Scientific Applications:

• APA profiling across protocols: Enables analysis of APA in PolyA-seq, PAC-seq, and other 3'Seq datasets.
• Discovery of novel polyadenylation sites: Detects APA sites that are missed by reference-based approaches.
• Characterization of complex APA events: Captures multi-isoform genes with more than two APA isoforms, including non-proximal to non-distal shifts.
• Disease and perturbation studies: Applied to glioblastoma NUDT21 knockdown data to reveal increased numbers of genes with APA changes.
• Tissue and cell-type APA evaluation: Applied to MAQC brain, UHR, mouse hippocampal, and human stem cell-derived neuron datasets for comparative APA analysis.

Methodology:

Abstracts genes into APA matrices, applies iterative consensus non-negative matrix factorization (NMF) based clustering, and uses vector projections to infer differential APA usage.