mRNALocater

mRNALocater predicts mRNA subcellular localization to support analysis of mRNA functional roles and cellular compartmentalization.

Key Features:

• Model Integration: mRNALocater employs a model fusion strategy that integrates XGBoost, LightGBM, and CatBoost models built using an optimal feature subset.
• Enhanced Prediction Accuracy: By leveraging multiple algorithms, mRNALocater achieves higher prediction accuracy compared to mRNALoc for certain subcellular localizations.
• Advanced Sequence Encoding: The tool encodes mRNA sequences using electron-ion interaction pseudopotential and pseudo k-tuple nucleotide composition to capture rich sequence information.
• Feature Mining Techniques: mRNALocater applies correlation coefficient filtering and feature forward search to mine and select informative features.
• Generalization Ability: The combined encoding, feature selection, and model fusion strategy improves generalization across diverse mRNA datasets.

Scientific Applications:

• mRNA localization mapping: Predicts subcellular localization of mRNAs to inform studies of intracellular RNA distribution.
• Gene expression regulation: Supports investigation of spatial regulation mechanisms affecting mRNA function.
• Cellular compartmentalization studies: Aids analysis of compartment-specific mRNA roles in cellular processes.
• Omics research support: Provides data useful for genomics, proteomics, and systems biology analyses.

Methodology:

mRNALocater encodes sequences with electron-ion interaction pseudopotential and pseudo k-tuple nucleotide composition, applies correlation coefficient filtering and feature forward search for feature selection, integrates XGBoost, LightGBM, and CatBoost via model fusion using an optimal feature subset, and has been validated on independent datasets.