Known sequence features can explain half of all human gene ends

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• Cleavage and polyadenylation (CPA) sites are crucial for defining eukaryotic gene termination.
• Key sequence elements (UGUA, PAS, U-rich, CA/UA, DSEs) are known to be involved in CPA site recognition.
• The sufficiency of these elements and the role of other factors in delineating CPA sites remain unclear.

Purpose of the Study:

• To investigate the sufficiency of known sequence features in predicting constitutive cleavage and polyadenylation sites.
• To determine the contribution of individual sequence elements and RNA-binding proteins (RBPs) in CPA site definition.
• To assess the predictive power of established features for human gene end identification.

Main Methods:

• Utilized standard discriminative models to analyze sequence features.
• Evaluated models based on the five primary CPA sequence elements.
• Incorporated U1-hybridizing sequences and known RBP RNA-binding motifs into the models.

Main Results:

• Models with only the five primary CPA sequence features accurately predicted constitutive CPA sites at gene ends for 41% of human genes.
• U1-hybridizing sequences offered a minor improvement in prediction performance.
• Adding known RBP RNA-binding motifs enhanced prediction accuracy to 49%, suggesting involvement of known and novel regulatory factors.

Conclusions:

• The five core sequence elements are highly effective in predicting constitutive CPA sites for a significant portion of human genes.
• RNA-binding proteins play a substantial role in refining CPA site selection and preventing cryptic site usage.
• This study demonstrates the strong predictive power of established features for identifying human gene ends, with potential for discovering new regulatory mechanisms.