The scientific community has gained a powerful tool for analyzing genetic information. The AIRI Research Institute has introduced the GENATATOR neural network platform, capable of automatically mapping genes based on DNA sequences. This is not just another model—it is a comprehensive solution that changes the approach to genome annotation.

How It Works

GENATATOR is a multi-stage AI system. It sequentially solves several tasks: first, it identifies probable transcript start and end points, then analyzes the regions between them, refines the internal structure of genes, and filters out questionable hypotheses. As a result, the system can identify both protein-coding genes and long non-coding RNAs—key elements in the regulation of cellular processes.

Advantage Over Classical Methods

The key difference between GENATATOR and traditional approaches is the abandonment of searching for individual markers (start and stop codons, splicing signals). Instead, the neural network learns from complete DNA sequences, identifying deep patterns. This approach allows the tool to work with genome assemblies of non-model organisms for which detailed annotations are lacking.

The model was trained on genes from humans and 38 other mammalian species, including exotic ones like walruses and elephants. However, testing showed impressive results on other organisms as well—fruit flies, Arabidopsis thaliana, and baker's yeast. In terms of accuracy, GENATATOR significantly surpassed existing analogs.

Application Prospects

The tool opens up broad possibilities: from studying evolutionary processes and searching for new genes to analyzing the genomes of ancient animals. In the medical field, GENATATOR can be used for predicting hereditary diseases, developing personalized medications, and biotechnological research.

The models are already publicly available on the Hugging Face platform. A web service and an open leaderboard for quality assessment have also been launched. Currently, the local version requires a CUDA GPU with float32 format—execution on a central processor and reduced-precision modes are not supported.

My analysis: GENATATOR is a significant step forward in the field of bioinformatics. The ability to annotate genomes without being tied to specific organisms will accelerate research in genetics and medicine. However, the dependence on expensive equipment (CUDA GPU) may limit accessibility for small laboratories. In the future, optimizing the model for CPU and cloud computing will make this tool truly widespread.