A hybrid residue based sequential encoding mechanism with XGBoost improved ensemble model for identifying...

Area of Science:

• Molecular Biology
• Bioinformatics
• Epigenetics

Background:

• RNA modifications are crucial for cellular regulation, influencing gene expression and protein function.
• Cytosine hydroxymethylation, mediated by TET enzyme oxidation, is a key RNA modification with significant epigenetic implications.
• Current methods for detecting 5-hydroxymethylcytosine (5hmC) are costly and time-consuming.

Purpose of the Study:

• To develop an efficient and accurate machine learning algorithm for identifying 5-hydroxymethylcytosine (5hmC) in RNA.
• To overcome the limitations of traditional laboratory methods for 5hmC detection.
• To enhance the interpretability of machine learning models in RNA modification analysis.

Main Methods:

• Proposed XGB5hmC, a machine learning model utilizing the XGBoost algorithm.
• Employed diverse residue-based formulation methods and fused six frequency residue-based encoding features into a hybrid vector.
• Integrated SHAP (Shapley Additive Explanations) for feature selection and model interpretability.

Main Results:

• The XGBoost ensemble model achieved superior performance compared to existing state-of-the-art methods.
• Achieved high accuracy (89.97%), sensitivity (87.78%), specificity (94.45%), F1-score (0.8934%), and MCC (0.8764%) using tenfold cross-validation.
• SHAP analysis identified key features contributing to the model's predictive power.

Conclusions:

• XGB5hmC provides a computationally efficient and accurate method for 5hmC identification.
• The model's interpretability enhances understanding of RNA modification mechanisms.
• This advancement holds potential for improving medical assessments and treatment strategies in RNA modification analysis.