Correlations between complex human phenotypes vary by genetic background, gender, and environment

A parametric bootstrap approach for computing confidence intervals for genetic correlations with application to genetically-determined protein-protein networks

“…Therefore, we have , meaning that the genetic variance is equal to the heritability. Under this assumption, the variance of the phenotype can be written as Given two n × 1 vectors y 1 , y 2 , their covariance can be modelled as where is the genetic variance for phenotype is the residual variance for phenotype i, ρ k is the genetic correlation between the two phenotypes, and ρ e is the residual correlation between the two phenotypes (15). Alternatively: If we further plug in , then, for a single and for two phenotypes, we can write the variance model as: which is the form that we will use to simulate outcomes in the following parametric bootstrap section.…”

“…While the procedure is in principle naïve to the specific formula used, we are using the closed-form Hasemen-Elson formulas we previously derived (15,20): Where W is either the kinship matrix with all diagonal values set to zero, or, a weighted sum of the kinship matrix K and the matrix modelling the random error (here, an identity matrix) with weights related to the relationship between the kinship matrix and the identity matrix. See (15) for more details, including the potential use of multiple matrices modelling correlations between individuals. In practice, it is appropriate to use the kinship matrix with diagonal value set to zero when only the kinship matrix is used to model relationship between individuals.…”

“…Methods for estimating heritability and genetic correlations based on summary statistics from GWAS (3,13,14) became popular in recent years due to their computational tractability and the access to many phenotypes that were interrogated in GWAS by the research community. However, in diverse populations and in small datasets it is still preferable to estimate heritabilities and genetic correlations using individual-level data, rather than based on GWAS summary statistics (15). Methods using individual-level data typically rely on an underlying linear mixed model (LMM) formulation, where a genetic relationship matrix is used to model the relationship, or degree of similarity, between the phenotype levels of different individuals (16,17).…”

“…When estimating genetic correlation between two phenotypes, a bivariate normal model is usually used. Common algorithms for estimating heritability and genetic correlation include Restricted Maximum Likelihood (REML)-based normal likelihood models (18), and method of moment estimators such as the Haseman-Elston approach (15). Estimation of standard errors (SEs), confidence intervals (CIs), and p-values, often relies on asymptotic normal approximation.…”

The Rise of Gastrointestinal Cancers as a Global Phenomenon: Unhealthy Behavior or Progress?

“…For such complex integrative tasks, it is possible that a multiomics-like approach, already showing a relevant role in cancer diagnosis, survival analysis, and response to treatment [ 201 , 202 ], may also be applied for investigating potentially new and unexplored associations underlying GI cancer pathogenesis. Multiomics approach has recently provided novel insights into the biological mechanisms behind gene–environment interactions [ 203 , 204 ]. In this line, a recent study with multiomics profiling identified several associations revealing potential biological responses and sources of exposure in early life, including signatures for diet, chemical compounds, trace elements, and weather conditions, among others [ 205 ].…”