Normalization of cDNA microarray data.

TLDR: The print-tip loess normalization as mentioned in this paper is a well-tested general purpose normalization method which has given good results on a wide range of arrays and can be refined by using quality weights for individual spots.

About: This article is published in Methods.The article was published on 2003-12-01 and is currently open access. It has received 2084 citations till now. The article focuses on the topics: Normalization (statistics).

Figures

Figure 2. Spatial plot of green background values. The array was printed using a 12 x 4 pattern of print tips. The image shows that the array tends to be more green around the edges in the four corners. There is also a green patch in tip rows 8 and 9 and columns 3 and 4.

Figure 6. Between array scale normalization for a series of six arrays. (a) Side-by-side boxplots of the M-values from 6 arrays. The arrays are replicates except that three are dye-swap pairs of the others. Array 5 has a much larger spread than the others. (b) Boxplots of the same arrays after scale-normalization to equalize the median absolute value for each array. Data from the Corcoran Lab, WEHI.

Figure 1. MA-plot showing three different trend lines. The horizontal blue line shows the median of the M-values. The continuous orange curve shows the overall trend line as estimated by loess regression. The yellow curve shows the loess curve through a set of control spots known to be not differentially expressed.

Figure 5. Plate or print-order effects for the first slide in the ApoAI knock-out experiment.

Figure 4. Individual loess curves for the 48 print tip groups for the same array as in Figures 1 to 3.

Figure 3.Boxplots of the M-values for each print tip group. The M-values are higher (more red) in the middle of each sequence of four and in the middle of the overall sequence.

Frequently asked questions

Q1. What contributions have the authors mentioned in the paper "Normalization of cdna microarray data" ?

This article describes normalization methods based on the fact that dye balance typically varies with spot intensity and with spatial position on the array. When diagnostic plots show that biases still remain in the data after normalization, further normalization steps such as plate-order normalization or scalenormalization between the arrays may be undertaken. 

Q2. What future works have the authors mentioned in the paper "Normalization of cdna microarray data" ?

Normalization methods for cDNA microarrays will no doubt see further development in the future, but print-tip loess normalization provides a well-tested general purpose normalization method which gives good results on a wide variety of arrays. The method may be refined by using quality weights for individual spots. When diagnostic plots show that biases still remain in the data after normalization, further normalization steps such as plate-order normalization or scale-normalization between the arrays may be undertaken. 

Q3. What packages are used for normalization of cDNA arrays?

The Bioconductor packages use the free statistical programming environment R. For normalization of cDNA arrays, the relevant packages are marrayNorm [8, 9] and limma. 

Q4. What is the prerequisite for weights to be useful?

The prerequisite for weights to be useful is that they should be numerical and inversely proportional to the variances of the M-values. 

Q5. Why do the authors not use 2D normalization as a routine method?

The authors do not use 2D normalization as a routine normalization strategy because of concern that imperfections on the array may present sudden rather than smooth changes and concern that the 2D loess curve may confuse local clusters of differential expression on the array with the spatial trend to be removed. 

Q6. What is the meaning of the command to normalize the M-values in the array?

The commands to identify non-differentially expressed control spots will depend on the naming conventions in the allocation list as well as on the detailed controls printed on the array, so it is impossible to give universally applicable commands. 

Q7. What is the way to normalize the MSP curve?

In the composite normalization procedure, it is best to use constant local regressions (degree=0) to construct the MSP loess curve so that any necessary extrapolation of the MSP curve outside of the intensity range of the control spots will also be constant, this being the most conservative extrapolation policy. 

Q8. How many spots were printed with DNA?

Indeed it turns out that spots with print orders between 169 and 252 were printed with DNA from a different library to the other spots. 

Q9. How many pixels are in an ideal circular spot?

Inspection of the TIFF images of arrays used in the examples in the article suggests that the area in pixels of an ideal circular spot on these arrays is about 165 pixels. 

Q10. What is the way to normalize a printtip?

further normalization should be applied only when diagnostic plots show strong evidence of the need for such normalization, as unnecessary estimation and removal of trends adds noise to the data. 

Q11. What is the way to use base-2 logarithms?

It is convenient to use base-2 logarithms for and so that is units of 2-fold change and is in units of 2-fold increase in brightness. 

Q12. What is the meaning of the weight function given in Figure 7?

When using the SPOT image analysis program, the authors have found it useful to weight spots according to their area in lieu of a more comprehensive measure of spot quality. 

Q13. What is the way to normalize the M-values?

Print-order normalization should be used when, as in this case, exploratory plots of the data reveal a substantial print-order effect in the M-values.