Microarray Best Practice
The following recommendations were collected by B070 members. All items were
thoroughly discussed and approved in a plenary meeting on August, Thu 30th (2007).
If something should be added or has changed, please contact Kurt.
- Array Production:
- There's two established surface chemistries in our lab: Epoxy (ask Andrea)
and Aminosilane (ask Linda or Rafa/Brazil). LORECS have been reported to be problematic on
Epoxy, not keeping round shape. For linkers ask Anette.
- Ask Andrea or Christoph which pins to use for nucleic acids, for proteins,
for antibodies. Pin dependencies can completely mess up your data! Read this .
- Spotters: Microgrid (for huge amounts of plates, up to 100 slides,
refillable, ask Rafa/Brazil or
Michi), Esi (easier to use, up
to 100 slides, ask Andrea), or the small one
in Anette's lab (non-contact piezo, low-throughput, ask Anette).
- Spot in douplicate (or single spots) for M-CHiPS analysis.
- Spot duplicates in different blocks/areas instead of next to each
other.
- If possible, spot enough "housekeeping" genes of different intensity
leves (ideally many intensity levels equidistant in log scale!) for your experimental context to
study for normalization (in case spiking does not work). There are no genes
constant under all possible conditions. Including 1/3 of the total number of
genes on the chip (scanned by Affymetrix analysis) will do. Also, adding 300
housekeeping genes to a total 7000 genes was successful.
- Spot enough different spiking controls (most probably a must for future publications).
Again, the more different intensity levels (ideally equidistant in log
scale) spiked in the better. For LORECS ask Jorge.
- Orientation matters! To touch your slide only at the label end while putting
it into the scanner, label as follows:
- MicroGridII for scanning with ScanArray 5000:
Slides are inserted vertically. Scratch the slide ID onto each slide at its
lower end (the end next to you) such that you can read the number (not upside-down!).
- MicroGridII for scanning with ScanArray 4000XL (the one with the
stacker):
Slides are inserted vertically. Scratch the slide ID onto each slide at its
top end (the end far from you) such that you can read the number (not upside-down!).
- Hybridization:
- Samples: Before processing a sample, see to it that enough data (sample
annotations, clinical data, ...) are available to characterize it (just the
profiles without meaning may not be sufficient for publishing).
- For experimental design, reference condition talk to Kurt.
- p-values: Permutation tests such as SAM need at least 6 (six
!)
hybridizations in each class (experimental condition, tumor type or the
like).
- Randomize application of chips and chip batches to experimental
conditions!! For how and why talk to Kurt.
- Scanning:
- Inserting - direction
- ScanArray 4000 with stacker: Move the slides into the metal slide rack
(stacker) AGAINST the arrow (i.e. NOT in array direction) in order not to damage the
small metal clips that hold the slide in place.
- ScanArray5000 (one-slide only): only one possible way.
- Inserting - orientation
Orientation matters! For both scanners, slides inside the scanner
MUST have exactly the same orientation as in the spotter when spotted
with MicroGridII (i.e. numbers readable, but that in ScanArray 4000 you
can't see them for being behind the stacker).
For Esi-spotted slides, make
sure to rotate correctly.
- You should scan with constant laserpower like with 70,
80, or, 90 but different PMTs.
For the analysis on GenePix you can use these
ones that "fit the best"
- or you may adjust each laserpower to the hybriziation during scanning
(much more time consuming)
- Save scanning in tif-format (Andrea)
- Imaging:
- Read through "Data example"
- Double-check orientation of the tif-image (rotate if necessary) before applying your
grid (wrong orientation assigns wrong names to all of your genes).
- Use Genepix Version 4.1 or 6.0 and save as Version 4.0! (Andrea, Frank)
- Do not flag misbehaving spots (will make things worse).
- Let Genepix adjust for different spot sizes but define a minimal spot
size (very important!).
- In case of background (e.g. disturbed by dust), move spot away. If
containing some signal, make it smaller (avoiding the dust) but not too
small (instable). See example.
- Additionally save the (manually adapted) grid for any .gpr file. That
is
- save results : gpr-file (numbers for mergeblocks and M-CHiPS)
- save settings: gps-file (the grid, to avoid repeating the whole imaging process e.g. if the .gpr file format is
wrong)
- Stop producing .gpr files until the first one is checked.
- Stop making hybridizations until you had a first look on the data.
- Analysis:
- Do not save any files that are to be kept permanently (e.g. .gpr) on spool
(deleted monthly or going to be deleted soon).
- You do not need to transfer any files, because all your windows folders
are visible from Unix/M-CHiPS. In case you encounter any problems, ask
Christoph, Marc, Christian.
- For M-CHiPS, prepare your database according to here and proceed like this.
- For each spot, the mean of its pixels is taken to analysis by default.
Reasons and alternatives are detailed here.