tRForest: a novel random forest-based algorithm for tRNA-derived fragment target prediction.

Area of Science:

• Molecular Biology
• Genetics
• Bioinformatics

Background:

• tRNA fragments (tRFs) are small non-coding RNAs with gene regulatory functions, similar to microRNAs (miRNAs).
• tRFs operate independently of Dicer, associate with Argonaute proteins, and post-transcriptionally regulate gene expression.
• Accurate prediction of tRF targets is crucial for understanding their biological roles, but existing methods have limitations in scope and feature selection.

Purpose of the Study:

• To improve the accuracy and scope of tRNA fragment (tRF) target prediction.
• To leverage established miRNA target prediction features within a machine learning framework for tRFs.
• To provide a validated computational tool for predicting tRF targets.

Main Methods:

• Applied a random forest machine learning algorithm using features established for miRNA target prediction.
• Utilized a comprehensive set of tRF classes for prediction.
• Validated predictions in two independent cell lines and performed Gene Ontology analysis.

Main Results:

• Achieved significant improvements in tRF target prediction across all tRF classes.
• Validated the predictive model's performance in independent experimental settings.
• Identified enrichment of predicted tRF targets in neuronal function pathways, particularly for tRF-3009a.

Conclusions:

• The developed tRForest approach substantially enhances tRF target prediction accuracy.
• Conserved tRFs, like tRF-3009a, play significant roles in neuronal function.
• These findings advance the understanding of tRFs' broad biological functions and offer a resource for future research.