FVTLDA

FVTLDA predicts potential associations between long non-coding RNAs (lncRNAs) and diseases using feature-vector-based transductive learning to support identification of biomarkers and therapeutic targets.

Key Features:

• Integration of Direct and Indirect Features: Uses feature vectors representing lncRNA-disease pairs and association probability fractions to integrate both direct and indirect information and enable predictions when no related lncRNAs or diseases are known.
• No Reliance on Negative Samples: Trains without requiring labeled negative samples, avoiding dependence on artificially constructed negatives.
• Combination with Multiple Linear Regression (MLR) and Artificial Neural Network (ANN): Applies MLR and ANN to refine FVTLDA predictions and enhance analytical capability.
• Cross-validation Performance: FVTLDA with MLR achieved AUCs of 0.8909 (fivefold CV), 0.8936 (tenfold CV), and 0.8970 (LOOCV), while FVTLDA with ANN achieved AUCs of 0.8766, 0.8830, and 0.8807 in the same settings, and average case study contrast scores were 0.8429 (MLR) and 0.8515 (ANN) versus 0.6375 for KATZLDA.

Scientific Applications:

• Gastric cancer case study: FVTLDA with MLR identified 8 of the top 10 candidate lncRNAs as verified by recent literature.
• Leukemia case study: FVTLDA with MLR identified 8 of the top 10 candidate lncRNAs as verified by recent literature.
• Lung cancer case study: FVTLDA with MLR identified 8 of the top 10 candidate lncRNAs as verified by recent literature.
• Biomarker and therapeutic-target prioritization: Ranks candidate lncRNAs to support discovery of novel biomarkers and therapeutic targets.

Methodology:

FVTLDA employs a transductive learning approach that integrates feature vectors and association probability fractions and uses Multiple Linear Regression (MLR) and an Artificial Neural Network (ANN) to refine predictions.