BiPSTP: Sequence feature encoding method for identifying different RNA modifications with bidirectional...

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• RNA modification is a key posttranscriptional regulatory mechanism impacting RNA function.
• Accurate identification of RNA modification sites is crucial for understanding biological roles.
• Existing RNA sequence encoding methods face limitations in information representation and feature integration.

Purpose of the Study:

• To develop a novel RNA sequence feature representation method, BiPSTP.
• To create an effective machine learning model, mRNAPred, for predicting RNA modification sites.
• To enhance the accuracy and robustness of RNA modification site identification.

Main Methods:

• Introduced BiPSTP (bidirectional trinucleotide position-specific propensities) for RNA sequence encoding.
• Utilized parameter ξ to capture positional and sequential information.
• Developed the mRNAPred model using a support vector machine classifier.

Main Results:

• The mRNAPred model significantly outperformed state-of-the-art models in identifying 12 types of RNA modification sites.
• The BiPSTP method improved the robustness and generalization performance of prediction models.
• BiPSTP showed potential for feature extraction from DNA sequences for predicting other biological modifications.

Conclusions:

• The BiPSTP method offers a superior approach to RNA sequence representation for modification site prediction.
• mRNAPred provides a robust and accurate tool for identifying multiple RNA modification sites.
• The BiPSTP method has broad applicability in predicting biological modification sites in both RNA and DNA.