NIA 15k Membrane Microarray Data Processing
Developmental Genomics & Aging Section, Laboratory of Genetics,
National Institute on Aging, NIH, Baltimore, MD 21224-6820, USA

 Original: Mark Carter, Ph.D  02/24/2003

Modified: Kazuhiro Aiba, Ph.D

Here is an overview of how we process our data from the membrane-based cDNA microarray system:

  1. Our microarray clone set is spotted on 7 separate filters (~2400 clones per filter), labeled A through G.

  2. Triplicate hybridizations are performed for each tissue being compared (so, for a standard A v. B comparison, that would be 6 hybridizations total), and the images are gridded and volume reports from Image Quant are generated. Boxes around the edge (numbers 17-34 in fig. 1) are positioned over non-DNA regions of the membrane, to be used for background subtraction.


    Figure 1 . one membrane filter, gridded

  3. Background Subtraction: On each filter, the volumes of all 18 background spots are averaged, and this average is subtracted from the volume of each DNA feature on the membrane.¹

  4. Normalization: For each hybridization, the background-subtracted volumes (intensities) of all DNA elements are summed across all seven membranes. This total is divided into an arbitrary maximum value (such as 1010 volume units), and the resulting normalization factor is used to multiply each background-subtracted value.²

  5. Student.s t-test: After background subtraction and normalization, we perform a standard student.s t-test on each DNA feature, at a confidence level of 95%. The means of the two samples are statistically significantly different if:
    This critical value depends on the sample sizes from which the two means are calculated, nB and nA, the significance level of the test (1% or 5%), the value of the standard deviation of the data, and the number of data points from which this standard deviation is calculated.
    where s is the calculated standard deviation. In our case, we calculate this as the .pooled. standard deviation of the two samples being compared:
    in which case the tCrit value has nA + nB . 2 degrees of freedom. See appendix A for student.s t-distribution table.



    Figure 2 . one membrane filter with new grid (blue)

Appendix A: Percentage Points, Student.s t-Distribution³

For one-tailed tests, use F = 1 . p value; for two-tailed tests, remember to use F = 1 . (p value / 2). n is the degrees of freedom used.

n\F .50 .75 .90 .95 .975 .99 .995 .9995
1 .325 1.000 3.078 6.314 12.706 31.821 63.657 636.619
2 .289 .816 1.886 2.920 4.303 6.965 9.925 31.598
3 .277 .765 1.683 2.353 3.182 4.541 5.841 12.924
4 .271 .741 1.533 2.132 2.776 3.747 4.604 8.610
5 .267 .727 1.476 2.015 2.571 3.365 4.032 6.869
6 .265 .718 1.440 1.943 2.447 3.143 3.707 5.959
7 .263 .711 1.415 1.895 2.365 2.998 3.499 5.408
8 .262 .706 1.397 1.860 2.306 2.896 3.355 5.041
9 .261 .703 1.383 1.833 2.262 2.821 3.250 4.781
10 .260 .700 1.372 1.812 2.228 2.764 3.169 4.587
11 .260 .697 1.363 1.796 2.201 2.718 3.106 4.437
12 .259 .695 1.356 1.782 2.179 2.681 3.055 4.318
13 .259 .694 1.350 1.771 2.160 2.650 3.012 4.221
14 .258 .692 1.345 1.761 2.145 2.624 2.977 4.140
15 .258 .691 1.341 1.753 2.131 2.602 2.947 4.073
16 .258 .690 1.337 1.746 2.120 2.583 2.921 4.015
17 .257 .689 1.333 1.740 2.110 2.567 2.898 3.965
18 .257 .688 1.330 1.734 2.101 2.552 2.878 3.922
19 .257 .688 1.328 1.729 2.093 2.539 2.861 3.883
20 .257 .687 1.325 1.725 2.086 2.528 2.845 3.850
21 .257 .686 1.323 1.721 2.080 2.518 2.831 3.819
22 .256 .686 1.321 1.717 2.074 2.508 2.819 3.792
23 .256 .685 1.319 1.714 2.069 2.500 2.807 3.767
24 .256 .685 1.318 1.711 2.064 2.492 2.797 3.745
25 .256 .684 1.316 1.708 2.060 2.485 2.787 3.725
26 .256 .684 1.315 1.706 2.056 2.479 2.779 3.707
27 .256 .684 1.314 1.703 2.052 2.473 2.771 3.690
28 .256 .683 1.313 1.701 2.048 2.467 2.763 3.674
29 .256 .683 1.311 1.699 2.045 2.462 2.756 3.659
30 .256 .683 1.310 1.697 2.042 2.457 2.750 3.646
40 .255 .681 1.303 1.684 2.021 2.423 2.704 3.551
60 .254 .679 1.296 1.671 2.000 2.390 2.660 3.460
120 .254 .677 1.289 1.658 1.980 2.358 2.617 3.373

¥

 .253 .674 1.282 1.645 1.960 2.326 2.576 3.291

¹In the time since our original data processing method was developed, we have tried to improve our background subtraction methods and algorithms. We considered possibilities such as adding more background measurements and interpolating background for each DNA feature based on position, calculating error using standard error propagation algorithms, and setting near-background intensity features to lower-limit values based on the variance of background measurements (Fig.2 above).

²While it may seem intuitive to be consistent about the levels at which we background subtract and normalize (i.e. If background subtraction is done at the membrane level, then normalization should also be done at the membrane level), we have found that normalizing at the membrane level can produce 2 to 7 .clusters. of data, presumably an artifact of intra-membrane differences in DNA content and background. In these cases, normalization at the hybridization level produces data which is not clustered by membrane location.

³This table is abridged from the .Statistical Tables. of R.A. Fisher and Frank Yates published by Oliver & Boyd, Ltd., Edinburgh and London, 1938.