How Should We Assess Glycemic Variability in Type 1 Diabetes? Contribution of Principal Component Analysis for Interstitial Glucose Indices in 142 Children.

BACKGROUND: Glycemic variability (GV) can be used to assess glycemic control in diabetes, but there is no clear consensus concerning the methods to use for its assessment. Methodological differences have resulted in differences in the outcome of GV metrics used in research studies, controversies over clinical impact, and an absence of integration into routine care. AIM: To identify the indicators of GV most meaningful for clinicians, patients, and clinical researchers. MATERIALS AND METHODS: Continuous glucose monitoring data were collected during the first 3 months of a pediatric diabetes clinical trial (Start-In!; n = 142). We used principal component analysis (PCA) to analyze weekly averages for 22 parameters relating to GV. RESULTS: PCA identified five groups of parameters and three components explaining 85.7% of the variance. These components represented the amplitude, direction (hypoglycemia vs. hyperglycemia), and timing (within-day vs. between-days) of glucose excursions. CONCLUSIONS: This study provides elements that could make GV parameters more useful in clinical practice and research. No single parameter was sufficient to represent the complexity of GV, but it was possible to restrict the number of indicators required. The five groups of parameters identified by PCA could facilitate the choice of the most relevant outcomes for GV analysis in pediatric diabetes according to the purpose of the analysis (e.g., exploration of GV associated with hypo- or hyperglycemia, with short- or long-term periodicity, or GV in its entirety).

Kernel Multitask Regression for Toxicogenetics.

The development of high-throughput in vitro assays to study quantitatively the toxicity of chemical compounds on genetically characterized human-derived cell lines paves the way to predictive toxicogenetics, where one would be able to predict the toxicity of any particular compound on any particular individual. In this paper we present a machine learning-based approach for that purpose, kernel multitask regression (KMR), which combines chemical characterizations of molecular compounds with genetic and transcriptomic characterizations of cell lines to predict the toxicity of a given compound on a given cell line. We demonstrate the relevance of the method on the recent DREAM8 Toxicogenetics challenge, where it ranked among the best state-of-the-art models, and discuss the importance of choosing good descriptors for cell lines and chemicals.