At the 'Artificial Intelligence in Science' conference, junior research fellow Nikita Trofimovich Lisenco presented a report on the use of AI in biology and medicine, highlighting achievements by Russian scientists. Scientists extended the life of E. coli 3.5 times using the Evo neural network pre-trained on billions of protein sequences. Modification of bacterial telomerase achieved this result.
The Evo and ESM3 neural networks helped decode genes responsible for yeast synthesis of the antibiotic teicoplanin. These genes were integrated into the genome of laboratory mice, which began to produce this substance, demonstrating resistance to bacterial infections without side effects.
The Institute of Spatial Proteomics Anton Kirillockichin and his team are working on the SigmaFold model to add a fourth dimension (time) in predicting protein structure, taking into account their dynamic formation process.
Russian scientists jointly with Slovenian researcher Matvey Korchwitznik developed a DNA decoding method using GigaChatPro and Gena-LM neural networks. The method uses arsenic substitution for phosphorus in DNA to create more stable DNA.
The Serbsky Institute of Psychiatry has launched a project to develop a pharmacological method for curbing deviant behavior, using Chilean degus rodents. Petr Belobelskiy's team applied machine learning for selecting new candidate substances.
The ChemBERTa–KANT model proposed new candidate substances for life extension that are already in the second stage of clinical trials both in Russia and abroad. Nikita Trofimovich Lisenco reported that the KANT idea was proposed by Jürgen Schmidhuber in 1991.