BATTER deep learning tool predicts bacterial transcription termination across 42,905 genomes (microbiomejournal.biomedcentral.com)

• BATTER uses deep learning to predict bacterial transcript 3′ ends from conserved stem-loop structures.
• Analysis of 42,905 genomes revealed clade-specific terminator properties and RUT-like sequences in Cyanobacteria.
• Tool identified widespread premature termination in AMR genes; open-source code and data provided.

"Researchers developed BATTER, a deep learning framework that predicts bacterial transcript 3′ ends by leveraging conserved stem-loop structures found in both Rho-independent and Rho-dependent terminators. Applied to 42,905 representative bacterial genomes, BATTER revealed clade-specific stem-loop properties, identified RUT-like sequences in Cyanobacteria lacking rho homologs, and systematically found premature termination events in antimicrobial resistance genes. The tool enables large-scale comparative genomics of transcription termination and is available as open-source software."