LncRTPred

LncRTPred predicts interactions between long non-coding RNAs (lncRNAs) and messenger RNAs (mRNAs) in human and mouse using machine learning to support functional interpretation of lncRNAs.

Key Features:

• Machine Learning Framework: Uses machine learning models rather than thermodynamic approaches and reports improved prediction accuracy compared to LncTar.
• Robust Negative Dataset Generation: Identifies non-interacting lncRNA-mRNA pairs from an unlabeled dataset using BLAT and filters these to obtain outliers for a reliable negative training set.
• Cumulative Model Score: Provides a cumulative_model_score for each query as a quantitative measure of interaction likelihood.
• Performance Evaluation: Evaluated against experimentally validated disease-specific lncRNA-mRNA interactions and shown to outperform other popular prediction tools in those tests.
• Species-specific Training Effects: Prediction performance is influenced by training dataset size, with better results reported for human due to a larger training set compared to mouse.

Scientific Applications:

• Functional Annotation of lncRNAs: Infers lncRNA functions by predicting their mRNA interaction partners.
• Study of Disease-specific Interactions: Identifies and prioritizes disease-specific lncRNA-mRNA interaction candidates for further study.
• Experimental Prioritization: Ranks candidate lncRNA-mRNA pairs using model scores to guide experimental validation.

Methodology:

Machine learning models trained on positive and negative datasets; negative dataset created by detecting non-interacting pairs from an unlabeled set using BLAT and filtering to obtain outliers; models evaluated against experimentally validated disease-specific lncRNA-mRNA interactions and produce a cumulative_model_score per query.