Dr. Iryna V. Lobach conducts research in applied statistics and supports the biostatistics component of biomedical research.

Her methodological research interests are motivated by challenges arising in the analyses of how an effect of the genetic basis varies by non-genetic measures (environment), what is traditionally referred to as gene-environment interaction (GxE) analyses. This type of analyses might provide valuable clues to the underlying pathophysiologic mechanisms of complex phenotypes.

For example, in collaboration with statisticians and biomedical researchers, Dr. Lobach developed an efficient retrospective analysis of case-control genetic studies in situations when the environmental variable is measured with error, e.g. diet or physical activity. Both in theory and in practice, this method improves efficiency of the risk parameter estimates and is capable of correcting coverage of 95% Confidence Intervals from e.g. 68% of the standard approach to the nominal 95% in a practical example of analyses of association between calcium intake and the risk of colorectal adenoma analyses.

We developed statistical methods to incorporate heterogeneity of the clinical diagnosis characterized by biomarker studies in situations when 1) multiple distinct pathophysiologic mechanisms share symptoms and hence the clinical diagnosis; and 2) substantial segments of the population are undiagnosed. We show that the genetic variant will appear to interact with the environmental variable if the genetic variant affects the pathologically defined disease state and the environmental variable is related to the proportion of cases with that disease state. We then propose a pseudo-likelihood solution and apply the methods to large genome-wide studies of Prostate Cancer and Alzheimer’s disease.

Recently, we developed case-only estimates of GxE in settings when 1) multiple disease state share the clinical diagnosis, 2) frequencies of the disease state of interest within the clinical diagnosis vary by the E or other variables, 3) both clinical diagnosis and the disease states are common. We show the application to Alzheimer's disease.

We are also conducting studies to investigate bias in the genetic effect estimates due to omitting a continuous variable. Specifically, we are interested to assess what information is needed to correct the bias if the actual values of the continuous variable are not available to the researcher. For example, many of the genetic databases offer only a brief set of non-genetic variables. We show the application to Alzheimer's disease.

Dr. Lobach supports the biostatistics component of multiple biomedical studies through collaboration, consultation (CTSI biostatistics consultant) and mentoring (K-grants, mentoring in the Training in Clinical Research Program).