GTA

GTA classifies gene transfer agent (GTA) genes and distinguishes them from viral homologs using a support-vector machine trained on amino acid composition to enable accurate identification of RcGTA-like elements in microbial genomes.

Key Features:

• Classification Capability: Distinguishes RcGTA-like genes from typical viral proteins by detecting differences in amino acid composition.
• Training Flexibility: Initially trained on 11 RcGTA genes from Rhodobacter capsulatus (Kogay et al., 2019) and can be retrained for other RcGTA genes, GTAs from different organisms, or other virus-like elements given appropriate training data and compositional separation.
• Implementation: Implemented in Python.
• Efficiency and Accuracy: Detected RcGTA-like "head-tail" gene clusters in 57.5% of 1,423 examined alphaproteobacterial genomes in benchmarking analyses.

Scientific Applications:

• Genomic Analysis: Enables annotation and detection of previously unrecognized RcGTA-like elements within alphaproteobacterial genomes.
• Prophage Misidentification Correction: Identifies cases where more than half of in silico prophage predictions may actually represent GTAs, informing correction of misannotations.
• Evolutionary Studies: Facilitates investigation of GTA involvement in horizontal gene transfer and evolutionary dynamics in bacterial populations.

Methodology:

Uses a support-vector machine classifier trained on manually curated datasets of homologous viruses and GTAs to classify genes based on amino acid composition, with the model configurable for retraining on different gene sets or organisms.