tRFTars

tRFTars predicts target genes of tRNA-derived fragments (tRFs) by modeling tRF–mRNA interactions to identify regulatory targets.

Key Features:

• Target prediction: Predicts human tRF target genes from tRNA-derived fragments (14–40 nucleotides) by modeling tRF–mRNA interactions.
• Training data: Utilizes CLASH (crosslinking, ligation, and sequencing of hybrids) and CLEAR-CLIP (covalent ligation of endogenous AGO-bound RNAs) datasets to identify tRF–mRNA interactions.
• Feature selection: Employs a genetic algorithm to select predictive features including minimum free folding energy (MFE), position 8 match, number of bases paired in the tRF–mRNA duplex, and tRF length.
• Model construction: Builds support vector machine (SVM) prediction models and reports performance with area under the ROC curve (AUC) values of 0.980 (training) and 0.847 (testing).
• Validation: Validates predictions with quantitative real-time PCR (qRT-PCR) overexpression assays showing down-regulation of mRNA levels for thirteen predicted target genes.

Scientific Applications:

• Gene regulation studies: Identification of tRF target genes to investigate tRF-mediated regulation of gene expression.
• Disease mechanism exploration: Analysis of tRF–mRNA interactions to provide insights into diseases associated with dysregulated RNA processes.
• Therapeutic target identification: Prioritization of genes influenced by tRFs as potential therapeutic targets in pathological conditions.

Methodology:

Models were trained on CLASH and CLEAR-CLIP interaction data, with feature selection performed by a genetic algorithm (features: MFE, position 8 match, number of paired bases, tRF length) and prediction performed using support vector machines (SVMs).