Guy Brock, PhD
Associate Professor
Department of Biomedical Informatics
Ohio State University
Presentation Title
Imputation for Truncated High Dimensional Data with Applications to Metabolomics Studies
Abstract
High throughput technology (e.g., mass spectrometry) makes it possible to monitor metabolites in biological experiments and has been widely used to detect differences in metabolite abundance between samples. A frequent issue encountered with these samples is missing values, which can arise from different sources including both technical and biological reasons. Often the missing value is substituted by the minimum value, and this substitution may lead to biased or suboptimal results in downstream analyses. In this talk I will discuss several methods we developed which specifically accounts for the truncation at the minimum value, and allows for a mixture of missingness due to missing at random (MAR) and missing not at random (MNAR) values. The first approach uses a modified version of the K-nearest neighbor (KNN) algorithm, called KNN truncation (KNN-TN). The second is a Bayesian model, called BayesMetab, that systematically accounts for missing values based on a Markov chain Monte Carlo (MCMC) algorithm that incorporates data augmentation by allowing MVs to be due to either truncation below the LOD or other technical reasons unrelated to its abundance. We compare these two approaches to other imputation algorithms based on a variety of performance metrics (power for detecting differential abundance, area under the curve, bias and MSE for parameter estimates). Applying our approach to an analysis of metabolomics data from a mouse myocardial infarction study revealed several statistically significant metabolites not previously identified that were of direct biological relevance.