Generative AI Bridges Gaps in DNA Microscopy Data to Advance Genetic Medicine

Generative AI Completes Missing Microscopy Data to Advance Genetic Medicine

Researchers at Skoltech have developed a method using generative artificial intelligence (AI) to fill gaps in microscopy data that measure distances between gene pairs in DNA. This breakthrough enables a clearer picture of DNA's 3D structure, which is vital for improving diagnostics and treatments of genetic diseases.

The study, published in Scientific Reports, represents the first successful attempt to reconstruct missing gene distance data using AI or any other method. Previously, scientists relied on incomplete datasets that limited progress in medical genetics and the biophysics of chromatin—the material that makes up chromosomes.

Why DNA’s 3D Architecture Matters

DNA function depends not just on the sequence of genes but also on their spatial arrangement. The folding of the 46 DNA molecules inside each cell affects gene activity and influences processes like cell reproduction and differentiation during embryonic development. Faulty DNA folding contributes to diseases including cancer.

Understanding the physical principles behind a healthy 3D DNA structure opens doors to new diagnostic markers and therapeutic targets. By comparing DNA structures in healthy and diseased cells, researchers can design drugs that restore normal gene function or develop precise gene editing techniques.

Limitations of Fluorescence Microscopy Data

Fluorescence microscopy is commonly used to study DNA folding by tagging specific gene sequences with fluorescent markers. However, this data is inherently incomplete. Some DNA sequences, especially those with repeated nucleobases, cannot be selectively stained, leading to missing distance measurements between certain gene pairs.

Until now, this missing data posed a significant obstacle. But generative AI provides a solution.

How AI Bridges the Data Gaps

Assistant Professor Kirill Polovnikov from Skoltech Neuro explains that once enough gene distance data is known, predicting the missing distances becomes a solvable mathematical problem with a unique solution. Their team demonstrated that generative AI models—usually applied to creative tasks like image or text generation—can solve this problem effectively.

This approach introduces a fresh angle to chromatin research, traditionally dominated by polymer physics.

Implications for Research and Medicine

• Practical: The method improves processing of fluorescence microscopy data, enabling more accurate mapping of DNA’s 3D structure. This advancement supports the development of better genetic disease diagnostics and therapies.
• Conceptual: The study highlights generative AI’s potential beyond typical creative applications, showing its usefulness in complex biological data reconstruction.

The research was funded by the Russian Science Foundation (Grant No. 25-13-00277).

For professionals interested in the intersection of AI and biological data analysis, exploring courses on generative AI models can be valuable. Resources like Complete AI Training offer tailored programs to deepen knowledge in AI applications.