April 29, 2019 at 15:58AI @ MIPT: Big data for mathematical models of the human genome
Hello, Habr! FizTech resumes a series of workshops on artificial intelligence AI @ MIPT. Today, Alexander Frey and Kevin O'Connell of the Norwegian Center for the Study of Psychological Disorders (NORMENT) in Oslo will talk about using big data for mathematical models of the human genome. The performance can be seen in the group of Fiztekh , it will be held in English.
Genetics is an area of great potential for machine learning. The most representative full genome studies (GWAS) are already conducted on samples exceeding 1 million people and contain information on tens of millions of associations between genomic variants and phenotypic traits. According to some estimates, in total, these studies cover 80% of the variability of complex traits of a person, including mental disorders. Despite this, the difficulty of moving from acquired scientific knowledge to clinical applications remains. In practice, personal genetic information is rarely used to predict disease. During the lecture, the most successful genome-wide studies and limitations that impede the effective use of machine learning methods in human genetics will be examined. Part of the lesson will be devoted to the statistical methodology underlying such studies, including the analysis of Bayesian mixed models and the method of limited maximum likelihood. The topics of polygenetic risk assessment and precision medicine, which is already being effectively implemented in terms of assessing the personal risk of cancer and stratifying the risks of developing Alzheimer's disease, will be touched upon.
Previous lectures of the course can be seen here .
Genetics is an area of great potential for machine learning. The most representative full genome studies (GWAS) are already conducted on samples exceeding 1 million people and contain information on tens of millions of associations between genomic variants and phenotypic traits. According to some estimates, in total, these studies cover 80% of the variability of complex traits of a person, including mental disorders. Despite this, the difficulty of moving from acquired scientific knowledge to clinical applications remains. In practice, personal genetic information is rarely used to predict disease. During the lecture, the most successful genome-wide studies and limitations that impede the effective use of machine learning methods in human genetics will be examined. Part of the lesson will be devoted to the statistical methodology underlying such studies, including the analysis of Bayesian mixed models and the method of limited maximum likelihood. The topics of polygenetic risk assessment and precision medicine, which is already being effectively implemented in terms of assessing the personal risk of cancer and stratifying the risks of developing Alzheimer's disease, will be touched upon.
Previous lectures of the course can be seen here .