Experimental Design for Microarray Analysis

The large-scale screening of thousands of genes at the same time by microarrays would be impossible without bioinformatics. Data analysis is becoming more and more challenging, with increasing complexity of experimental design and biological systems. Large-scale screening of genes requires careful experimental design, as there are some facts for which bioinformatics cannot compensate. The final goal needs to be taken into account. For example, a diagnostic tool monitoring a severe infection or even sepsis will certainly need much higher confidence levels and thus higher sample numbers than a random screen for possibly interesting genes. The exploration of a single pathway on a DNA chip (e.g., pentose phosphate cycle in the human pathogen Listeria monocytogenes when switching from extracellular to intracellular survival) with a limited number of genes may make global normalization impossible; a sufficient number of independent sequences from house keeping genes and external controls need to be spotted on the array

1.7 Adaptation in Time and to Stimuli | 9

for signal intensities to be adjusted. In dual labeling experiments, if more than two samples are to be compared, it is useful to run each sample against some internal standard so as to be able to compare relative signal intensities. This internal standard might be some control RNA closely related to the test samples or a commercially available reference RNA. To introduce as little variation as possible, a sufficient amount of reference RNA for all arrays planned in one study should be prepared, pooled, and aliquoted at the beginning of a study. To avoid labeling specific false positive signals, it is essential to include dye switch experiments if two probes are being compared on the same slide. Pooling of several probes may cancel out real differences and at the same time increase the false positive rate due to dominating outliers. Finally, different hybridization protocols will give significantly differing results, and therefore protocols, including possible RNA amplification, must not be changed during the experiment.

0 0

Post a comment