Understanding how genes are organized and regulated is a big part of studying how organisms function. In bacteria, many genes are grouped together into units called operons, which are transcribed as a single unit. These gene clusters often control related biological processes, and figuring out which genes belong to the same operon can provide important clues about their function.

Researchers from Memorial University of Newfoundland have developed a powerful new tool called OpDetect to identify operons more accurately. Unlike earlier tools that only work with a limited number of bacterial species or require specific types of data, OpDetect works across a wide variety of organisms. It uses RNA sequencing (RNA-seq) data and applies advanced deep learning methods, including convolutional and recurrent neural networks, to read patterns directly from the genome.

Feature generation process

The read counts per nucleotide (nt) for Gene 1, intergenic region (IGR) and Gene 2 are proportionally sampled into a 150-feature vector. This vector is separated into three channels, one for each region. On each separated vector, the features outside the corresponding region are zeroed. If there is not IGR, the corresponding channel will be all zeros.

OpDetect stood out for its performance, outperforming previous tools in key metrics like recall, F1-score, and AUROC. Impressively, it even works on Caenorhabditis elegans, a rare case of a eukaryotic organism that also contains operons. With its broad species coverage and open-access code available on GitHub, OpDetect is a versatile and promising addition to the bioinformatics toolkit.