In control: systematic assessment of microarray performance

Abstract

Expression profiling using DNA microarrays is a powerful technique that is widely used in the life sciences. How reliable are microarray-derived measurements? The assessment of performance is challenging because of the complicated nature of microarray experiments and the many different technology platforms. There is a mounting call for standards to be introduced, and this review addresses some of the issues that are involved. Two important characteristics of performance are accuracy and precision. The assessment of these factors can be either for the purpose of technology optimization or for the evaluation of individual microarray hybridizations. Microarray performance has been evaluated by at least four approaches in the past. Here, we argue that external RNA controls offer the most versatile system for determining performance and describe how such standards could be implemented. Other uses of external controls are discussed, along with the importance of probe sequence availability and the quantification of labelled material.

Figure 1

Accuracy and precision. (A) A method is accurate and precise when it repeatedly returns a measurement close to the real value. (B) If a method contains a systematic error, it might frequently return an identical measurement that is lower than the actual value; this allegation is frequently made against microarray data (Yuen et al, 2002). In such a case, the measurement is precise but inaccurate. (C) If measurements suffer from noise, the average of a series of measurements might still return the real value but with a large standard deviation; in this case, the measurement is accurate but not precise. (D) The worst case is when measurements report the incorrect value with a large standard deviation.

Figure 2

Assessing spiked ratios. A self-versus-self hybridization with nine external control RNAs that are spiked in at different ratios. The average expression ratio from a merged dye-swap experiment is plotted as a function of the average background-subtracted intensities of the two channels (R and G). The two channels were lowess-normalized on genes per print-tip (Yang et al, 2002b). Genes are indicated in grey. Ratio controls were spiked twofold up (red) or down (green). One control was added in equal amounts to both channels (yellow). Each control is represented at least 96 times on the arrays that are used, which results in the different clusters of control spots. The spread of each cluster is dependent on both hybridization and spotting-pin uniformity.

Figure 3

Dose–response test. A log–log plot of the signal intensity versus concentration for 11 control RNAs. Reproduced with permission from Lockhart et al (1996) Nature Biotechnology © Nature Publishing Group.

Figure 4

Quantification of labelled material. Absorption spectra for Cy3- and Cy5-labelled cDNA after the removal of unincorporated dye. Peaks for cDNA, and incorporated Cy3 and Cy5, are found at 260, 550 and 649 nm, respectively. Two mock-labelled samples are included as negative controls in which reverse transcriptase was left out of the cDNA-synthesis reaction. This is important for determining the success of purification and the removal of RNA template by hydrolysis. The cDNA yields and dye incorporation can be calculated using the indicated formulas.

Frank C. P. Holstege & Harm van Bakel. FCPH is the recipient of an EMBO Young Investigator Award