Prediction of uridine modifications in tRNA sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• Limited research exists on post-transcriptional modifications, specifically tRNA modifications, despite their role in genome structure and codon usage.
• Uridine modifications (UMs) are a major class of tRNA modifications, making them a key focus for understanding these processes.
• This study aims to explore kingdom-wise tRNA modifications, with a particular emphasis on UMs.

Purpose of the Study:

• To develop an efficient and accurate computational method for predicting kingdom-wise tRNA uridine modifications (UMs).
• To evaluate the performance of prediction models across different kingdoms and datasets.
• To classify prominent UMs, such as Pseudouridine (Y) and Dihydrouridine (D).

Main Methods:

• A three-step strategy was employed, starting with a common prediction model for all kingdoms using Support Vector Machine (SVM).
• A hybrid approach combining binary and structural information yielded the highest performance (AUC 0.936).
• Kingdom-wise prediction models were developed and evaluated using independent datasets from MODOMICS-2012.

Main Results:

• The common prediction model showed high performance across kingdoms (AUC 0.910–0.949) on independent datasets.
• Individual kingdom-wise models achieved even higher prediction accuracy (AUC 0.915–0.987).
• The hybrid approach proved effective for predicting and classifying major UMs (Y and D).

Conclusions:

• The developed hybrid approach is efficient for predicting kingdom-wise tRNA modifications.
• The method successfully classifies prominent UMs into Pseudouridine (Y) and Dihydrouridine (D).
• A user-friendly webserver, tRNAmod, has been created for predicting UMs from tRNA sequences and whole genomes.