Genetic 'skylines' in the aetiology of graded pharmaceutical phenotypes

Humans show variation in pharmaceutical response (whether that be in safety, efficacy, quality of life measures or health costs) due to both their genetics and their environment. This ’pharmacogenetics’ and ’pharmacoepidemiology’ respectively, can be characterised in any one individual as a multidimensional pattern of measurements - a profile - within a specific context. Inferences can be drawn from comparing sets of profile sizes or ’skylines’, together with the examination of outlier individuals within such groups, according to the question being addressed. Bowman (2009) outlines how a research question to be answered can be captured as a data transformation using individualised divergences when comparing pharmacogenetic profiles from case-control studies. Mathematical insights from such contrasts can inform the understanding of the biological mechanisms involved and the clinical aetiology of the response. Bowman (2013) offers an entry point into how unordered profiles have been manipulated using individualised divergences in different pharmacogenetic contexts to date. Profile sizes or ’skylines’ have been investigated by:- summarisation over measures or over individuals; by network visualisation; and, by examination of the eigen decomposition of correlations within them. Procrustean formalism in shape space has not yet been pursued. One challenge remains as to how to deal with comparative ’skylines’ when there may be an ordered phenotypic gradient between groups. Bowman (2015) poses combining new individualised divergences based upon scaled orthogonal quadratic contrasts to augment existing two-group linear forms as a solution to detect in-homogenous signals. Two real data examples explaining and validating this proposal for insights into the size changes of comparative genetic ’skylines’ along three-point phenotypes in drug adverse reactions is offered.