MicroArrayGroupInYahoo Some thoughts on validation of microarray slides (RickTearle, Tue Jul 9 03:11:03 2002)
From: http://groups.yahoo.com/group/microarray/message/6956
The standard of validation of microarray slides is fairly poor. When you buy a slide you really have no idea if every spot is present and whether it will hybridise efficiently. Below we make some suggestions of how users might go about improving the validation of slides and experiments.
Labelling the oligo
One obvious method of determining what is on a slide is to label the spotted [DNA], which can then be scanned to generate an image of the slide. However a barrier to this approach is the added cost of appending a fluor to an oligo (it can double the initial cost). So instead, just label one in every thousand oligos.
This can be done by using a 1000:1 mixture of CPG and eg fluorescein-CPG to start the oligo synthesis. Using so little FluoresceinCpg per oligo will have virtually no impact on the cost of the oligo.
The signal from such an oligo will also be much weaker than if every oligo is labelled and energy transference to a bound cyanine labelled DNA strand becomes much less likely. Ideally, it would be best to adjust the level of fluorescein labelling of the spotted DNA so that it fluoresces with roughly the same magnitude as the hybridised cyanine labelled [DNA]. You then have the option of scanning the three fluors after hybridisation and aligning all three images, with the possibility of correlating the cyanine signals to the fluorescein signals. Missing or badly spotted spots will become obvious straight away.
What about if you don't have a scanner capable of scanning three fluors? Well, if you are getting your slides from a third party, ask them to scan them for you. If you had different scanners it will be difficult to align the three images, but at least you will know what is on the slide.
Purifying the oligo?
Because of coupling failures during synthesis a significant proportion of long oligos are not full length. Does this matter? We have seen no data on this point, but we suspect that spotting only full length oligos is more likely to give consistent data when comparing samples which are polymorphic. However purifying long oligos is a costly exercise and again will increase overall costs. An alternative is to add a linker to the 5' end of the oligo. Only full length oligos will carry this linker, as failed oligos cannot be added to. It was interesting to read BartJanssen's summary of his experiments with different slides and chemistries, and the message we took away was that clean glass slides may be the best surface to spot onto.
One linking method we would be interested to follow up is EpoxyLabel-ling of the oligos for fixation to glass (contact ShishirShah at SpectralGenomics for more information on this). Regardless of the method chosen, we see value in adjusting the volume of the oligo so that the concentration of full length oligo is the same across the oligo set. This will of course require some method of quantitating the 5' linker and have not addressed this at all.
How much DNA to spot?
If we assume that we spot long oligos with a 5' linker, and the [DNA] only attaches via this linker, we can work out the area of the spot, the area of the DNA and thus get an upper limit of how much to spot. A double stranded DNA molecule occupies about 1nm2 in cross section and the area of a 100탆 diameter circular spot is about 7800탆2. So the maximum number of molecules you would want to bind is about 7.9 x 10^9 molecules, which is about 13fmole or 220pg of a single stranded 50mer.
Of course, not all the DNA may bind and you may want to use more to ensure efficient binding, but this figure gives a ball park of what to aim for.
Testing the oligos
One of the hardest aspects to validate is that the oligos work! For large genomes it is difficult to test large oligo sets. To overcome this, we suggest the following as a cost effective solution:
- take the oligos in sets of a thousand at a time (or 3x 384 well plates);
- ligate a 10-15 base sequence to the 3' end of the pooled oligos;
- hybridise a 10-15 base complement to the ligated 10 base sequence;
- extend the 10 base complement using Taq polymerase;
- cycle a number of times to generate a labelled sedcond strand in excess to the template strand;
end label the second strand with a fluor and hybridise to the MicroArray.
It will become clear which oligos fail to hybridise effectively and they can be retested and resynthesised.
We can envisage modifications of this method where the first strand is tethered to a StreptavidinMicroplate via a BiotinMoiety, making it easy to both purify the second strand and to reuse the template. It may also be possible to synthesise 10-15 base complement with a fluor already attached at the 5' end, obviating the need to label after second strand synthesis, but we are unsure as to what fluors are capable of withstanding ThermoCycling.
In theory, it would be possible to choose a fourth fluor for this control material and to hybridise it along side Cy3 and Cy5 labelled material each time a slide is used, this guaranteeing the efficiency of hybridisation across a slide is consistent from slide to slide. {For comparison of four fluors on the one slide, white light/CCD scanners with their capacity to hold the area under illumination constant while rotating in and out the filters for mulitple fluor quantitation, may offer real advantages.}
We put forward these suggestions in the hope that they may be useful to MicroArray users. The current level of slide validation is close to non-existent and surely must be improved in the interests of accuracy and reproducibility of MicroArray data.
RickTearle AndrewJones a1-biotech UK