Podium Presentation in Room 2 on Wednesday at 9:00 (Chair: Melanie Odenkirk)
Authors: Lisa Ford, Anne Evans, Douglas Toal
Introduction: In research settings, metabolomic experiments require statistically-powered comparisons between subject groups (e.g., diseased groups versus healthy groups) to identify changes in biochemical pathway measurements. Recently we have been interested in applying metabolomic principles in a clinical setting wherein individual patient samples (i.e., N-of-1) are analyzed against a previously run reference population to screen for inborn errors of metabolism. In this setting, data for hundreds of analytes in the patient sample are returned allowing for the comprehensive identification of perturbed metabolic pathways. Using consistent normalizing samples, the clinical metabolomics data are returned using z-scores to indicate the relative difference of each analyte versus the median response of the analyte in the reference population. Quality control procedures used in clinical metabolomics must encompass sample preparation, instrument analysis, and overall consistency of data for every analyte reported, and must allow data to be reported in a timely fashion.
Objectives: Design a comprehensive quality control strategy consistent with best practices in metabolomics and clinical analysis.
Methods: Internal standards are spiked at the sample level, both before extraction and during reconstitution in order to evaluate the precision of these steps. Duplicate quality control samples at two different levels are added to each batch and each analyte evaluated for precision against a set median and standard deviation based on historical data. The number of analytes found is also evaluated against historical data. Sample data are also mined for analytes indicative of decreased sample quality.
Results: The multilayer QC process begins with evaluation of the response of the internal standards used to monitor extraction and instrumental analysis, performed immediately following instrumental analysis. The internal standards typically perform within a 10% RSD, and any samples deviating from these criteria are further investigated. Normalized data for each analyte are generated, and the normalized data in the quality control samples are compared back to the set limits based on the median and standard deviations. By evaluating the number of analytes outside 2 and 3 standard deviations in the quality control samples, as well as evaluating the number of compounds found, it is possible to identify the conditions in which the assay is not performing in control. Semi-automated tools are used to identify and track the number of analytes found in the batch, the number and identity of the analytes outside acceptance criteria in the quality control samples, and sample quality parameters.
Conclusion: Though a daunting task due to the number of analytes, samples, and number of parameters to be examined, we have successfully developed a quality control scheme to aid in the delivery of consistent, high-quality, clinical metabolomics data.
|Planning to mention or discuss specific products or technology of the company(ies) listed above:||