Russian Scientists Develop AI Model to “Read” DNA Chains and Target Disease
Researchers have applied artificial intelligence to decode G-quadruplex structures, opening new paths for diagnosing and treating severe illnesses.
Scientists at the National Research University Higher School of Economics have developed an AI model that could expand the toolkit for diagnosing and treating serious diseases. The researchers used artificial intelligence to study G-quadruplexes – DNA structures that strongly influence how cells function and how organs and tissues develop. The university’s press service told IT-Russia about the work.
DeepGQ for G-Quadruplexes
As the researchers explain, DNA can be thought of as a long chain made up of four letters: A, C, G, and T. What matters is not only their sequence, but also how the strand is folded. Some regions are open, allowing the cell to read and replicate genetic information, while others are closed. There is also another form of “packaging” known as a G-quadruplex – a small knot-like structure formed in regions rich in the letter G, or guanine. Scientists have long suggested that each cell type has its own unique set of these knots that helps define its role. For example, neurons in the brain differ from liver cells in their G-quadruplex patterns, and those differences shape how various cell types develop. Detailed study of this phenomenon, however, has been limited because traditional methods are expensive and not always precise.
The Moscow HSE team says it has found a solution. The researchers developed an AI model called DeepGQ, which creates tissue-specific maps of G-quadruplexes. The model analyzes DNA sequences in two directions at once, reading them from left to right and right to left, allowing it to accurately capture the full pattern of features in a given DNA segment.
DeepGQ-Patient
As a result, any laboratory studying conditions such as liver cancer or Alzheimer’s disease can take patient sample data and, using DeepGQ, generate a precise map of the most likely targets for further analysis.
The researchers believe this work could eventually lead to a “DeepGQ-Patient” model. Based on a tumor biopsy and its molecular data, scientists could generate a personalized map of active G-quadruplexes and select a strictly individualized treatment strategy.
Earlier, IT-Russia reported that artificial intelligence in Russia is also being trained to predict the progression of diseases, although deploying such systems at scale is expected to take at least five years.
