[Genetic risk assessment of the joint effect of several genes: Critical appraisal]

When assessing the combined action of genes on the quantitative or qualitative phenotype we encounter a phenomenon that could be named the “paradox of the risk score summation.” It arises when the search of risk allele and assessment of their combined action are performed with the same single dataset. Too often such methodological error occurs when calculating the so called genetic risk score (GRS), which refers to the total number of alleles associated with the disease. Examples from numerous published genetic association studies are considered in which the claimed statistically significant effects can be attributed to the “risk score summation paradox.” In the second section of the review we discuss the current modifications of multiple regression analysis addressed to the so called “n ≪ p problem” (the number of points is much smaller than the number of possible predictors). Various algorithms for the model selection (searching the significant predictor combinations) are considered, beginning from the common marginal screening of the “top” predictors to LASSO and other modern algorithms of compressed sensing.