LncLocFormer

LncLocFormer predicts the subcellular localization(s) of long non-coding RNAs (lncRNAs) using a transformer-based deep learning model to identify localization patterns and sequence motifs relevant to multi-label localization assignment.

Key Features:

• Transformer Architecture: Employs eight Transformer blocks to model long-range dependencies within lncRNA sequences.
• Localization-Specific Attention Mechanism: Utilizes a localization-specific attention mechanism to discern relationships between different subcellular localizations and capture compartment-specific signals.
• Multi-Label Prediction Capability: Performs multi-label prediction to assign multiple subcellular localizations to a single lncRNA sequence.
• Motif Analysis Integration: Integrates motif analysis to capture known sequence motifs indicative of specific localization patterns.

Scientific Applications:

• Functional inference of lncRNAs: Predicts localization to support hypotheses about lncRNA cellular function.
• Multi-compartment localization studies: Provides comprehensive profiles of lncRNA distribution across cellular compartments via simultaneous multi-label predictions.
• Investigation of cellular processes and diseases: Facilitates studies of lncRNA involvement in cellular processes and disease mechanisms through localization insights.

Methodology:

The model analyzes lncRNA sequences with a transformer-based architecture comprising eight Transformer blocks, applies a localization-specific attention mechanism and motif analysis, and outputs multi-label subcellular localization predictions.