The Advances in Deep Learning Modeling of Polyadenylation Codes.

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• 3 -end cleavage and polyadenylation (polyA) is crucial for eukaryotic gene expression.
• PolyA site formation relies on variable, combinatorially acting sequence motifs.
• Quantifying polyA activities and defining cleavage sites presents technical challenges due to motif variability.

Purpose of the Study:

• To review advances in deep learning models for polyadenylation regulation.
• To discuss the applications of these models in understanding polyA site formation and function.

Main Methods:

• Deep learning models are employed to analyze sequence complexity and motif interactions.
• These models predict cleavage probability and quantify polyA site strength.
• The review summarizes existing literature on deep learning applications in polyadenylation research.

Main Results:

• Deep learning models can capture complex positional interactions among cis-regulatory motifs.
• These models provide novel insights into species-specific polyA site configurations and cleavage heterogeneity.
• Applications include analyzing genomic parameters and human genetic variants affecting polyadenylation.

Conclusions:

• Deep learning has significantly advanced the study of polyadenylation regulation.
• These models offer powerful tools for dissecting sequence complexity and regulatory mechanisms.
• Future research can leverage these methods to further elucidate polyadenylation biology.