Modeling of Spatial Distribution of Knockout Effect for Genes Associated With Aggressiveness of Low Grade Glioma in Human Brain Tissues Using Machine Learning

Currently, new experimental methods are widely used in the field of transcriptomic data analysis aimed at investigating the expression patterns of genes in different tissues under the influence of various environmental and internal factors, including polymorphisms. In particular, existing methods of gene knock-out and knock-down allow modeling of the impact that external factors have on expression of a target gene. Making use of available data on gene expression in diverse parts of an organism, including different brain regions, provides basis for constructing statistical models for mutual interdependencies of gene expression levels. Allen Brain Atlas database, for example, contains unique data on spatial distribution of gene expression levels in human and mice brain tissues. For the first time, the approach of mathematical modeling of spatial distribution of gene knock-out effects in human brain tissues using machine learning methods and data on gene expression taken from the Allen Brain Atlas database was suggested. It is shown that knock-out of central genes of gene network related to the aggressiveness of low grade glioma has stronger influence on expression of other genes in comparison with knock-out of peripheral genes in this network. Moreover, the effect reveals pronounced spatial heterogeneity.