Deep Learning Prediction of Inflammatory Inducing Protein Coding mRNA in P. gingivalis Released Outer Membrane...

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Porphyromonas gingivalis outer membrane vesicles (OMVs) are crucial in periodontitis pathogenesis.
• Identifying protein-coding mRNA sequences within OMVs is vital for understanding their function.
• Current methods for mRNA prediction in OMVs require enhancement.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting protein-coding mRNA sequences in P. gingivalis OMVs.
• To compare the performance of Naïve Bayes, Neural Networks, and Gradient Boosting algorithms for this task.
• To identify the most effective model for accurate and reliable mRNA sequence prediction.

Main Methods:

• Transcriptomic data from P. gingivalis OMVs were collected and annotated.
• Machine learning tools, including Orange, were used for data preprocessing and analysis.
• Naïve Bayes, Neural Networks, and Gradient Boosting models were trained and validated using cross-validation and ROC curves.
• Model performance was assessed using metrics such as AUC, Classification Accuracy, and F1 Score.

Main Results:

• Gradient Boosting demonstrated a balanced performance with an AUC of 0.72, Classification Accuracy of 0.41, and F1 Score of 0.32.
• Neural Networks achieved an AUC of 0.721, Classification Accuracy of 0.391, and F1 Score of 0.314.
• Naive Bayes showed lower performance with an AUC of 0.701, Classification Accuracy of 0.172, and F1 Score of 0.114.
• While no significant statistical differences were found, Gradient Boosting offered a more balanced precision-recall relationship.

Conclusions:

• In silico analysis using machine learning effectively predicted protein-coding mRNA sequences in P. gingivalis OMVs.
• Gradient Boosting emerged as a promising tool, outperforming Naive Bayes and Neural Networks in achieving balanced predictive performance.
• The findings suggest Gradient Boosting's potential as a reliable method for identifying protein-coding mRNA sequences in bacterial OMVs.