Biostatistics

Research at NCTR

Novel Methodology Supporting Rapid Detection of Bacterial Food Contaminants – National Center for Toxicological Research (NCTR) mathematicians have developed a novel algorithm to predict the optimal testing range and then identify characteristic bacterial genes found in comparative genomic microarray (“gene chip”) technologies that are proposed for rapid detection of food contamination. The novel statistical technology, known as change-point estimation, relies upon finding an optimal graphical representation from multiple data points in noisy systems (multivariate adaptive regression spines). It has outperformed existing algorithms for specificity and accuracy in the analysis of two publically available data sets. Comparative genomic analysis for bacterial contaminants in foods requires more advanced statistical techniques because of a high probability that food samples may simultaneously contain several different species.

Critical Path – New methods to evaluate clinical outcomes – The biostatisticians in the Division of Personalized Nutrition and Medicine, NCTR, have developed a novel classification tool (i.e., classification algorithm) to address the statistical challenge that arises in most clinical studies where huge sets of measurements are gathered on relatively few clinical trial participants. If not carefully considered, analysis of high-dimensional data of this type can easily generate models that do not predict the correct treatment strategies. The newly developed algorithm was used to analyze genomic data sets obtained from lymphoma patients and lung cancer patients to distinguish disease subtypes for optimal treatment and to analyze genomic data obtained from breast cancer patients to identify patients most likely to benefit from adjuvant chemotherapy after surgery. The performance of the proposed algorithm is consistently highly ranked compared to the other classification algorithms. A description of this algorithm has been published in the journal Artificial Intelligence in Medicine (Vol. 41, pp. 197–297, 2007).

The statistical classification method for individualized treatment of diseases developed is expected to play a critical role in developing safer and more effective therapies that replace one-size-fits-all treatments.