Prediction of non-coding RNAs in Fusobacterium nucleatum-infected mice using machine learning.

Area of Science:

• Microbiology and Bioinformatics
• Genomics and Transcriptomics
• Computational Biology and Machine Learning

Background:

• Fusobacterium nucleatum is prevalent in periodontal disease, yet its non-coding RNA (ncRNA) regulation mechanisms remain understudied.
• Previous genomic and transcriptomic studies offer limited predictive power for identifying ncRNAs.
• Understanding F. nucleatum's role in disease progression requires investigating its ncRNA regulation.

Purpose of the Study:

• To predict previously uncharacterized non-coding RNAs (ncRNAs) in F. nucleatum-infected mice using machine learning (ML).
• To identify long non-coding RNAs (lncRNAs) and circular RNAs (circRNAs) associated with periodontitis.
• To evaluate the efficacy of different ML algorithms for ncRNA classification.

Main Methods:

• Utilized a periodontitis gene expression dataset (GSE225589) from the GEO database.
• Identified and labeled long non-coding RNAs (lncRNAs) and circular RNAs (circRNAs).
• Applied and compared three machine learning algorithms: random forest (RF), adaptive boosting (AdaBoost), and naïve Bayes (NB) for classification.

Main Results:

• Random forest (RF) and adaptive boosting (AdaBoost) models demonstrated superior performance in classifying lncRNAs and circRNAs.
• Both RF and AdaBoost achieved a perfect area under the ROC curve (AUC) of 100%.
• The naïve Bayes (NB) model achieved a slightly lower AUC of 92%.

Conclusions:

• This study pioneers the application of machine learning for predicting ncRNAs in F. nucleatum-infected mice using transcriptomic data.
• RF and AdaBoost algorithms show high potential for accurately identifying infection-associated lncRNAs and circRNAs.
• Further validation with larger cohorts and external datasets is recommended to confirm these predictive findings.