TACOS

TACOS predicts the cell-specific subcellular localization of long noncoding RNAs (lncRNAs) across 10 diverse human cell types to inform functional inference based on localization.

Key Features:

• Tree-based stacking approach: Integrates multiple tree-based classifiers in a stacking ensemble to improve prediction accuracy.
• Six tree-based classifiers: Employs six different tree-based classifiers as base learners within the stacking framework.
• Ten feature descriptors: Uses ten distinct feature descriptors tailored for each cell type to represent lncRNA sequences.
• Cell-specific prediction: Predicts subcellular localization for lncRNAs within 10 diverse human cell types.
• Balanced training datasets: Constructs balanced training datasets specific to each cell type to support robust model training.
• Performance evaluation: Evaluates models using comprehensive cross-validation and independent assessments.
• AdaBoost integration: Incorporates AdaBoost baseline models and selects a tree-based classifier for final predictions.
• Comparative performance: Demonstrated consistent superiority over other methods tested in the study.

Scientific Applications:

• lncRNA localization annotation: Enables computational annotation of lncRNA subcellular locations across cell types.
• Functional inference: Supports inference of lncRNA molecular functions and mechanisms that depend on cellular localization.
• Cellular process studies: Facilitates research into processes influenced by lncRNA localization, such as cell cycle regulation and genome rearrangements.

Methodology:

Applies a tree-based stacking approach that integrates six tree-based classifiers and AdaBoost baseline models, uses ten feature descriptors per cell type, trains on balanced cell-specific datasets, and evaluates via cross-validation and independent assessments.