Find the 'most susceptible individuals' from the 'rarest mutations', and AI accurately predicts the pathogenicity of genetic variants

Typically, nearly everyone carries dozens of potentially harmful rare variants. One of the greatest difficulties in clinical assessment of common variants is the inability to effectively identify individuals who may be at high risk for developing the disease. In the research of the Primate Genome Project, scientists used the artificial intelligence neural network PrimateAI-3D to locate highly pathogenic rare mutations through evolutionary analysis with the idea of ​​"using the rarest mutations to find the individuals most susceptible to disease". and used to predict an individual's risk of disease.

The team of Professor Tomás Marquez-Bonet from Spain and the Illumina Artificial Intelligence Laboratory jointly conducted multiple research groups to compare the whole genome sequencing data of a total of 809 samples of 233 primate species and identify 4.3 million genetic mutation sites on human orthologous proteins that may lead to changes in protein structure were identified.

The researchers used the above-mentioned genetic mutation sites as the basis of the data set, added them to human disease genetic data, and trained the PrimateAI-3D artificial intelligence neural network with a gene data set containing 4.5 million types of possible benign mutations, so that it can More accurately predict the pathogenicity of genetic variants.

Genetic variation is one of the main causes of disease. Based on the genetic relationship between nonhuman primates and humans, the same genetic mutations may lead to similar results, and mutations that are common in primates may mean that these mutations are more likely to be harmless or have extremely limited harm.

So, how to predict the risk of a person’s genetic factors for common diseases such as diabetes and cardiovascular disease? Is it better to evaluate as the sum of thousands of common genetic variants that have a small effect, or as the sum of a few rare mutations that have a significant effect?

Comprehensive research shows that common and rare variants have complementary roles in predicting human disease risk. Common mutations identify more individuals who are likely to have the disease, while rare mutations make it easier to find the abnormal individuals at highest risk. Therefore, including rare variants in clinical assessments may be better than using only common variants in identifying the extreme individuals who are ultimately affected by most diseases and who are most in need of treatment or suffer from severe early-stage disease. It is of great significance for preventive screening.

This study successfully demonstrates the application of combining primate population sequencing data with deep learning models, which helps us understand the pathogenicity of human genetic variants and can help personalized genomic medicine provide better clinical results. Diagnostic guidance.

Author: Collective creation and compilation by the research group of Professor Zhang Guojie of Zhejiang University

Editor: Xu Qimin

*Wenhui exclusive manuscript, please indicate the source when reprinting.

The above is the detailed content of Find the 'most susceptible individuals' from the 'rarest mutations', and AI accurately predicts the pathogenicity of genetic variants. For more information, please follow other related articles on the PHP Chinese website!