Researchers at the AIRI Institute Advance GENA for Viral and Bacterial Detection in Bat Droppings
Researchers from the AIRI Institute for Artificial Intelligence are exploring new capabilities for the GENA neural network, focusing on its ability to detect viruses and bacteria within bat guano. The work is led by Olga Kardymon, a bioinformatician who heads AIRI’s Bioinformatics group, and is shared with the wider research community through socialbites.ca.
The team is actively developing additional methods for applying GENA. One key objective is to leverage metagenomic sequencing data from bat droppings to assess whether GENA can distinguish the various viral and bacterial genomes present in this complex environmental sample. This line of inquiry seeks to determine if the model can accurately identify and categorize multiple genomes captured in a single litter, a task that challenges conventional sequencing analysis.
GENA stands out as the first language model for DNA trained on the most complete version of the human genome. It was released to the public at the end of March 2022. The scope of its potential applications is broad: geneticists can use it to study how mutations influence gene function, explore different regions of the genome, and classify organisms based on sequencing data, among other tasks.
Kardymon emphasizes that GENA highlights regions of biological importance. Even when a particular genomic region lacks established connections in the scientific literature to specific functions, this does not imply an error in the neural network. Instead, it may indicate undiscovered biological significance. Such findings warrant further investigation and could gradually illuminate the mysteries of the human genome.
The article also invites readers to learn more about how neural networks search for mutations in the human genome, generate novel proteins, and predict the effectiveness of vaccines and drugs, with additional material available on socialbites.ca and attributed sources.