Генеративный ИИ от российских учёных может ускорить анализ ДНК

Russian researchers have used neural networks to accurately determine the distances between genes, which in the future will simplify drug development and disease diagnostics

Scientists from Skoltech have developed an algorithm based on generative artificial intelligence that can restore missing data on the location of genes in DNA. This method allows for highly accurate determination of distances between genes even with incomplete initial data, which can significantly accelerate research in the field of genetics and medicine.

As explained by Kirill Polovnikov, a senior lecturer at Skoltech, the task of determining the distances between genes based on a limited set of data boils down to a mathematical problem that can now be solved using generative models.

If you find out the distances between a sufficient number of pairs of genes, finding the distances between the remaining pairs, for which there is no experimental data, takes the form of a mathematical problem with a specific solution. We have shown for the first time that such problems can be solved by generative models, which is an atypical area for their application.
Kirill Polovnikov, senior lecturer at Skoltech

The human genome consists of long strands of DNA packed into chromosomes. Understanding their three-dimensional structure is important for studying the work of genes and their impact on the development of diseases. Traditional methods, such as fluorescence microscopy (FISH), do not always allow obtaining complete data due to technical limitations. Russian scientists have proposed using generative AI to fill these gaps, similar to how neural networks complete images.

During the experiment, the DDRM algorithm successfully restored the missing data on a section of human chromosome 21, surpassing classical methods. This opens up prospects for faster and more accurate genome analysis, which is critical for the development of new drugs and methods for diagnosing hereditary diseases.

Read more materials on the topic:

A method for accelerated fine-tuning of neural networks was presented by researchers from HSE and AIRI

The Russian platform for quantum chemical modeling was presented by Rostec at the CIPR-2025 conference

A new word in non-invasive cancer therapy: special nanoparticles were created by scientists from MII IM SFU

Sources
TASS

Now on home