THRONE

THRONE predicts human RNA N7-methylguanosine (m7G) modification sites to enable accurate identification of m7G locations for studies of mRNA processing, translation, export, and splicing.

Key Features:

• Ensemble learning architecture: A three-layer ensemble that integrates four distinct computational frameworks for m7G site prediction.
• First layer: Constructs 54 baseline models that generate initial predicted probabilities for m7G sites.
• Second layer: Builds six meta-models that use the baseline models' predicted probabilities as new feature vectors.
• Third layer: Employs a random forest super learner that integrates novel features derived from previous layers to produce final predictions.
• Sequence-based feature set: Leverages a wide array of sequence-based features across machine-learning classifiers.
• Systematic feature synthesis: Incorporates novel features at each ensemble stage to refine and enhance predictive performance.

Scientific Applications:

• m7G site identification: Identifying putative N7-methylguanosine modification sites in human RNA.
• mRNA biology studies: Facilitating formulation and testing of hypotheses about mRNA export, translation, and splicing related to m7G.
• Epitranscriptomics and gene regulation: Enabling investigation of the regulatory roles of m7G modifications in gene expression.
• Predictive benchmarking: Serving as a basis for performance assessment using cross-validation and independent evaluation of m7G predictors.

Methodology:

Three-layer ensemble comprising 54 baseline models, six meta-models that use baseline predicted probabilities as features, and a random forest super learner, combined with a wide array of sequence-based features; integrates four distinct computational frameworks and systematically synthesizes novel features at each layer.