Identification of D Modification Sites Using a Random Forest Model Based on Nucleotide Chemical Properties

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Dihydrouridine (D) is a crucial RNA modification found across life.
• D modification sites are implicated in cancer treatments.
• Current methods for D detection are inefficient and lack computational support.

Purpose of the Study:

• To develop an accurate and efficient computational model for identifying dihydrouridine (D) modification sites in RNA.
• To overcome limitations of traditional experimental techniques and existing computational tools.

Main Methods:

• Utilized eleven sequence-derived feature extraction methods.
• Implemented five popular machine learning algorithms.
• Employed oversampling techniques to address class imbalance during data preprocessing.

Main Results:

• The optimal model combined random forest with nucleotide chemical property modeling.
• Achieved high sensitivity (0.9688) and specificity (0.9706) in independent tests.
• Outperformed existing computational tools in independent validation.

Conclusions:

• The developed model offers a robust and reliable method for D site detection.
• This tool can significantly advance research into the functional roles of dihydrouridine.
• Facilitates further exploration of D modifications in biological and medical contexts.