Translation. Region: Russian Federal
Source: State University “Higher School of Economics” –
An important disclaimer is at the bottom of this article.
INInstitute of Artificial Intelligence and Digital Sciences FKN HSE University has proposed a new approach based on modern machine learning methods to determine a person's genetic origin. Graph neural networks make it possible to distinguish even very close populations with high accuracy.
Genetic analysis is a service that has become popular in the last 10-15 years not only as a medical diagnostic tool, but also as an opportunity to learn more about one's origins. DNA analysis allows one to assess ethnic composition, determine where ancestors lived and moved, and find the number of Neanderthal mutations in the genome.
This has become possible thanks to the development of modern technologies – genotyping, data storage and processing systems, machine learning – and a significant reduction in their cost. But at the same time, existing testing methods do not allow us to separate genetically close, related populations that have lived in adjacent territories for a long time.
Researchers at the HSE Institute of AI and Digital Sciences have developed a method that allows one to distinguish the origins of people from closely related populations. The technology is based on graph neural networks. The algorithm relies not on the DNA sequence itself, but on graphs that indicate genetic connections between people with common sections of the genome. Such sections reflect the degree of kinship between people and indicate how many generations ago they had common ancestors. The more matches, the closer the people are in origin. The vertices in the model correspond to a person, and the edges reflect the degree of kinship.
The method was tested on data from different regions. The results for the population of the East European Plain, for which a large database has already been collected, were especially interesting. The graph neural network was able to accurately determine the population affiliation of representatives of genetically very close peoples.
"Existing methods of genetic analysis solve a different problem: they determine belonging to large isolated populations, for example, they determine who had French, who had Germans, who had English in their ancestry. Our method allows us to work with closely related populations, which is especially relevant for Russia, a historically multinational country," says Alexey Shmelev, one of the authors of the work, a research intern.International Laboratory of Statistical and Computational GenomicsInstitute of AI and Digital Sciences, Faculty of Computer Science, National Research University Higher School of Economics.
In the future, the researchers plan to teach the neural network to predict the percentage of different populations in the genome.
The researchers registered theirdevelopmentcalled AncestryGNN — "Neural Network Prediction of Population Belonging from Common Genome Segments."
As Vladimir Shchur, head of the International Laboratory of Statistical and Computational Genomics at the Institute of AI and Digital Sciences of the Faculty of Computer Science at the National Research University Higher School of Economics, noted, the proposed method opens up new prospects for more accurately determining the population history of people and can be used in genealogical research and anthropology.
Please note: This information is raw content obtained directly from the source of the information. It is an accurate report of what the source claims and does not necessarily reflect the position of MIL-OSI or its clients.
.