SDLDA

SDLDA predicts associations between long non-coding RNAs (lncRNAs) and diseases by integrating linear and non-linear feature extraction using singular value decomposition (SVD) and deep learning.

Key Features:

• Hybrid Methodology: Integrates SVD and deep learning to capture linear and non-linear relationships in lncRNA-disease data.
• Linear Feature Extraction: Uses singular value decomposition (SVD) for dimensionality reduction and linear feature construction.
• Non-linear Feature Extraction: Employs deep learning models to learn complex, non-linear patterns and interactions.
• Predictive Evaluation: Demonstrates performance using leave-one-out cross-validation (LOOCV).
• Literature-based Validation: Validated through case studies with 28 of 30 cancer-related lncRNAs confirmed by biomedical literature across gastric, colon, and renal cancers.

Scientific Applications:

• Disease Mechanism Exploration: Infers lncRNA associations to aid interpretation of molecular mechanisms in diseases, particularly cancers.
• Biomarker Discovery: Identifies candidate lncRNA biomarkers for disease diagnosis and stratification.
• Experimental Prioritization: Provides computational hypotheses to reduce experimental screening burden for lncRNA-disease associations.

Methodology:

Data preprocessing of provided datasets; linear features extracted via SVD; non-linear features learned with deep learning models; combined features used to train the SDLDA model; performance assessed by leave-one-out cross-validation.