A two‐stage framework for generalized eukaryotic promoter prediction and activity classification
Year: 2026
Type: article
Abstract: Promoter-Net performs generalized eukaryotic promoter recognition and species-contextual promoter activity classification. Stage 1 learns promoter-associated sequence patterns across multiple eukaryotic species and is evaluated on an entirely withheld species. Stage 2 uses Functional Annotation of the Mammalian Genome 5 (FANTOM5) Cap Analysis of Gene Expression (CAGE) activity to distinguish high- and low-activity promoters within six vertebrates. Cross-validation, pooled versus species-specific testing, sequence-diversity analyses, motif enrichment, and surrogate-convolutional neural network (CNN) SHapley Additive exPlanations (SHAP) interpretation support model evaluation. The public web server and repository provide a reproducible workflow from sequence input to promoter and activity prediction. The web server is available at https://promoter-prediction-app.onrender.com/
Source: iMetaOmics.
DOI: 10.1002/imo2.70154
Authors: Muhammad Shujaat, Shi‐Qing Mao
Institution: Shenzhen University
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: diamond
OpenAlex ID: W7225635381