MP07:Tsiper:Multi-parametric Data Reduction of Phenotypic Screening for Quantitative Systems Pharmacology
Maria V. Tsiper1, Jennifer Sturgis1, Larisa V. Avramova1, Ray Fatig1, Shilpa Parakh1, Bartek Rajwa1, J. Paul Robinson1,3, V. Jo Davisson1,2
1Bindley Bioscience Center, Purdue University Discovery Park, 2Department of Medicinal Chemistry and Molecular Pharmacology, Purdue University Cytometry Laboratories; 3Department of Basic Medical Sciences and the Department of Biomedical Engineering, Purdue University
Early assessment and evaluation of new drug entities for mitochondrial dysfunction is now a critical feature for drug development. The timely detection of undesired mitochondrial toxicity before entry into the clinical trials and/or markets would benefit from high content screens capable of predicting levels of associated risks. Multi-parametric high content screening (mp-HCS) of mitochondrial toxicity holds promise as a lead in vitro strategy for drug safety evaluation. The increasing knowledge regarding factors controlling mitochondrial regulation and apoptosis further enhances the potential availability of biomarkers for mitochondria functional status. Cytometric technology offers superb platforms for producing high numbers of quantitative phenotypic parameters. A unique metric set for further automated evaluation using platform-independent data reduction and analysis offers additional avenues for comparisons of different biological systems with a wide range of candidate drug agents. In this study, we have developed a data driven statistical model for assessing differential responses for induced mitochondrial toxicity. Using a compound training set that covers a range of known pharmacological and toxicological activities, the mp-HCS measurements are demonstrated as robust enough to allow for quantitative comparisons of biological systems. The data analysis approach used in these screens enable incorporation of features due to variable responses from altered growth media such as glucose reduction. These additional biological response features establish higher discrimination of phenotypes due to drug-induced changes in mitochondrial and cellular responses. An example involves the use of galactose substitution for glucose to reveal compound features in cells that are sensitized to the drug actions. Specialized Cell Response-to-Induced Toxicity (SCRIT) vectors are developed to describe each compound and incorporates all features associated with differential response of cellular parameters including doses, cell types, and growth conditions. The dimensionality of SCRIT vectors depend on the number of chosen parameters which in turn depends on the hypothesis being tested. An example will be presented for comparative changes in dose responses for just 3 cellular function parameters. Using 84 training compounds, an automation approach is used for grouping of compounds according to mitochondrial involvement. Inclusion of 6 parametric responses enabled the discrimination of more subtle differences between compound toxicities within a common therapeutic class; scoring enabled a ranking of statins which were in direct agreement with clinical outcomes. Our results demonstrate that the mp-HCS approach using varied experimental conditions for feature extraction offers a basis for model creation and subsequent hypothesis testing for predicting levels of clinical risk. Comparison of drug-induced changes requires variations in glucose for suitable separation of mitochondrial dysfunction from other types of cytotoxicity. The number and choices of parameters used in the analysis and their statistical methods are fundamental for the specific hypothesis testing and the degree of phenotypic differences motivating new metrics in screen development.
|Click [+] for other articles on||The Market Place for Lab Automation & Screening||The Market Place|