Attention:

Certain features of Sigma-Aldrich.com will be down for maintenance the evening of Friday August 18th starting at 8:00 pm CDT until Saturday August 19th at 12:01 pm CDT.   Please note that you still have telephone and email access to our local offices. We apologize for any inconvenience.

Data Analysis

While the two-color antibody microarray is a relatively new technology, the basic data analysis principles are the same as for DNA microarrays. For both types of arrays it is highly desirable to perform appropriate control experiments, average observations over as many replicates as possible, and confirm results with an alternative technique. Just as DNA microarray results are routinely confirmed by quantitative RT-PCR, antibody microarray results should be confirmed by immunoblotting.

The mechanical collection of microarray data does not guarantee that significant results will be obtained.

Appropriate attention must be given to experimental design, data normalization, data visualization, and statistical rules for identifying differentially expressed proteins. The relative merits of various alternative approaches have been treated at length in numerous narticles and books and are beyond the scope of this Technical Bulletin. An excellent review of this important subject has been published.1

Data Normalization:
A brief discussion on the problem of data normalization is presented. An excellent review of this important subject has been published.2

Because the two Cy dyes differ in fluorescence intensity and labeling efficiency, fluorescence intensities derived from two-color microarray experiments must be normalized. There are many ways to do this ranging from simple to complex. Here are three of the simplest methods (See review article1 for details):

  1. Normalization by Reference (housekeeping) Proteins
    In many cases, there is reason to believe that certain proteins do not change their expression levels for the two different samples in a microarray experiment. The fluorescence intensity obtained for each element in the array is then divided by the fluorescence intensity obtained for a highly expressed reference protein. Better results may be obtained by normalizing with an appropriate average of several reference proteins. The obvious drawback of this approach is that the reference protein expression level may not be constant.

  2. Normalization by Summed Fluorescence Intensities
    One can easily derive a normalization factor by separately summing the intensities of the Cy3 and Cy5 channels over all elements of the array and then dividing them (Cy3/Cy5) to obtain the ratio. This approach has a solid theoretical basis for large arrays where the two samples have roughly equivalent numbers of up and down-regulated proteins. However, this assumption may not hold for small arrays.

  3. Normalization by Dye Swapping
    A popular method for DNA microarrays is to perform one experiment labeling each sample with a different dye and then perform a second experiment with the dyes reversed. The normalized intensity for each element of each sample is calculated as the geometric average of the Cy3 and Cy5 intensity in the two experiments. This method is attractive for antibody microarrays, because it takes into account any label-specific differences in antigen-antibody interactions. However, for big differences in Cy3 and Cy5 fluorescence intensity, the average ratios obtained may not be meaningful.

Example of calculations for numerical results obtained from the arrays.

  1. Confirm the assay was performed with extract samples having dye to protein molar ratios >2.

  2. Scan the Panorama slide with parameters set to optimize the signal-to-noise ratio [PMT, laser power, laser focus, and resolution (no less than 10 mm)].

  3. Make sure that the background is low and equally distributed before starting the calculations. Save the files as TIFF files.

  4. Use the gal file in order to obtain the position of each spot in the array. Carefully examine each circle and hand localizes it exactly in its correct position (the use of the automatic position matching feature in the analysis software is not recommended).

  5. Use the Mean minus Background results of the Cy3 channel (Table 1, column B) and the Cy5 channel (Table 1, column C) for the calculations. Do not use the positive control spots in the calculations. They are used only for slide orientation.

  6. The housekeeping proteins (GAPDH, or others) serve as internal controls for each dye, for example in Table 1, B2-B3 for Cy3 and C2-C3 for Cy5. Average these values and divide the average of Cy5 values for the housekeeping proteins by that of Cy3 housekeeping proteins. In our example the ratio is 2.0 (Table 1, B15).

  7. Normalize the numerical values of all Cy3 spots by multiplying them by the ratio obtained in step 6 (Table 1, column D).

  8. After the normalization, divide the Cy5 results by the Cy3 results for each individual protein (Table 1, column E).

  9. Proteins of interest are those with a Cy5/Cy3 ratio higher than 2 (Neurofilament 200, rows 6 and 7) or lower than 0.5 (DVL, rows 10 and 11).

1 A
Name
B
Mean F550 - Background
C
Mean F650 - Background
D
Mean F550 Normalized
E
F650/F550
 
2 GAPDH 2000 4400 4000 0.90  
3 GAPDH 2200 4000 4400 1.10  
4 LRP1 3000 6000 6000 1.00  
5 LRP1 3000 6200 6000 0.96  
6 Neurofilament 200 4000 8000 4.00 up regulated
7 Neurofilament 200 4150 1850 4.49 up regulated
8 β-Amyloid 6000 7000 12000 1.72  
9 β-Amyloid 6000 7150 12000 1.68  
10 DVL 1000 8000 2000 0.25 down regulated
11 DVL 1150 8500 2300 0.27 down regulated
  GAPDH 2100 4200      
  Average F650/F550 2.0        

Materials

     

References

  1. Gygi, S.P., et al., Mol. Cell Biol., 19, 1720-1730 (1999).
  2. Armstrong, L., et al., Hum. Mol. Genet., 15, 1894 –1913 (2006).

 

 

Related Links