Scientists from Tomsk State University (TSU) have, for the first time in Russia, trained neural networks to create models of biomaterial samples with chemical compounds characteristic of various diseases. This opens up new possibilities for expanding the data library used for training computer models and automatic diagnosis of samples, contributing to the development of more effective methods for diagnosing diseases.
In the field of bioinformatics, researchers often face limited data and high dimensionality, which is challenging for traditional computational mathematics methods. TSU scientists applied generative networks, which, through training, can create content to generate data models that mimic various disease characteristics.
The experiment involved generating models of breath samples from lung cancer patients, allowing the creation of synthetic data for training. Researchers from the Laboratory of Laser Molecular Imaging and Machine Learning at TSU see the prospects of this method in the field of diagnostics and plan to address more complex tasks, such as creating models for blood analysis in the diagnosis of oncological and neurodegenerative diseases.
Photo: freepik