| High throughput "omic" technologies, such as the cDNA microarray, have the potential of increasing mechanistic understanding of the biological underpinnings related to a biological outcome, enhancing safety assessments during the development of a new chemical entities, identification of new druggable targets, selection of patient candidates for therapeutic treatment, and for monitoring exposures to hazardous chemicals through biomarkers. However, for these potentials to be realized, investigators must ensure their experiments are properly designed with respect to their intended purpose, the data is appropriately managed to decrease human error, and prevent loss of data, and that the data are of sufficient quality to ensure the results are appropriate. To address these needs, the dbZach System, a database and associated computational applications, has been developed to manage data derived from toxicogenomic and pharmacogenomics experiments. Using historical data within the laboratory, a quality control protocol was developed, consisting of three different divisions. The first division uses a trained support vector machine (SVM), a statistical learning theory method, for identifying high and low quality arrays based on global intensity characteristics. The second division uses a semiparametric normalization method for identifying misaligned subgrids on the microarray, to ensure proper feature alignment and quantification. The third division utilizes boxplots to identify arrays with incongruent distributions, and line plots to identify trends with regards to the number of identified and saturated features. Using data within dbZach, three temporal experimental designs were compared: the independent reference, loop, and modified loop designs. By comparing the results from these experiments based on the amount of experimental error, identifying temporal confounds, and analyzing differences in the temporal clustering relationships, the modified loop design was judged the most appropriate design. However, when economic considerations are made, the loop design may be preferred when used with a larger number of biological replicates. |
