tRNA-DL

tRNA-DL applies deep neural networks to distinguish true transfer RNA (tRNA) genes from false positives in genomic sequences, thereby improving tRNA annotation accuracy.

Key Features:

• Deep Neural Networks: employs a deep neural network architecture to identify complex sequence patterns without handcrafted features.
• Training Data: trains on positive samples from known tRNA sequences and negative samples derived from false-positive predictions by tRNAscan-SE in coding regions.
• Feature Extraction: encodes input sequences using one-hot encoding to automatically generate features for model training.
• Performance Metrics: evaluated against existing methods and shown to reduce false positives when used alongside tRNAscan-SE.

Scientific Applications:

• tRNA annotation refinement: complements tRNAscan-SE in large-scale genomic studies to refine tRNA gene annotations and reduce erroneous predictions.

Methodology:

Known tRNA sequences were used as positive samples and false-positive predictions from tRNAscan-SE in coding regions as negative samples; sequences were one-hot encoded and input to a deep neural network architecture, and models were evaluated by comparing performance metrics and false-positive rates to existing methods.