Quality control for microarray analysis of human brain samples: The impact of postmortem factors, RNA characteristics, and histopathology

Abstract

The quality of results from microarray studies depends on RNA quality, which can be significantly influenced by postmortem factors. The aim of this study was to determine which postmortem factors and/or RNA electropherogram characteristics best correspond to microarray output and can be used to prospectively screen RNA prior to microarray analysis. Total RNA was extracted (N=125) from gray and white matter of postmortem frontal and occipital lobe tissue, acquired from normal controls, and patients with schizophrenia, bipolar disorder or major depression. Electropherograms were generated by the Agilent BioAnalyzer 2100, allowing calculation of the 28S/18S ratio, the 18S/baseline peak ratio and the RNA Integrity Number (RIN). These values were compared to post-hybridization image analysis of Affymetrix microarrays. The postmortem variables correlated with some quality measures but could not be used as effective screening tools. Logistic regression demonstrated that all three electropherogram measures were predictive for microarray quality, and that the RIN threshold predictive of "good quality" (>35% present calls) was most consistent with that of prior studies. The optimal RIN must be determined by the investigator's specifications for false inclusion and false exclusion. In contrast to RIN, the quality threshold for the 28S/18S ratio has proven unacceptably variable, due to sensitivity to slight differences in protocol and/or tissue source. In conclusion, the measures we found useful as screening criteria do not replace the need to exclude samples after a microarray analysis is performed, as an acceptable percent call rate and other measures of microarray quality represent the desired endpoint.