In the trend viewer under 'Groups' and then 'Define Signature' there are
already several options.
It would be nice if in this window you can also choose for: Signal > BG +
2SD in at least 1 Experiment.
In the .gpr (Genepix) files that we load, there is already a colum with
information about this.
It is the " % > B1+2SD " colum, which gives the percentage of pixels in a
spot that has more signal than the background plus 2 standard deviations
I hope this can help you to find a solution.
Best regards, Ellen
I was actually just going over an idea that a colleague had for your
question. His message is below. It is kind of a long workaround depending
on how many experiments are in the trend but it should give you the answer
you are looking for.
Unfortunately, the " % > B1+2SD " column from the GenePix file is not parsed
into Resolver because GenePix is the only Technology Type that uses it. I
will enter a feature request for some way to filter on intensity.
Have a look at the strategy below and let me know if it is practical for you
Here's what I found out in 3.2.2: the filter options still aren't good
enough to do what she's asking here--we get log(ratio), log(error), and
p-value. log(error) might give her an alternate way to get where she wants
to be, but maybe not.
But there is a workaround that isn't too odious for a smallish number of
experiments (say 10 or less).
1) Open the trend
2) Open one of the experiments which contributes to the trend. Go to
Groups->Filter and define the filter as you want it (probably "Intensity 1
less than X" or "Intensity 2 less than X").
3) Close the experiment. This saves your preferences, including the filter
4) Individually open all of the experiments which contributed to the trend.
5) Windows->Tile Windows (not technically necessary, but it makes the next
6) In the first individual experiment window, do Edit->Select Visible
followed by Edit->Broadcast Selection.
7) In the rest of the individual experiment windows, do
, - Groups->Selection->Hide other
, - Edit->Clear Selection
, - Edit->Select Visible
, - Edit->Broadcast selection
8) In the trend window, do Groups->Selection->Hide Others.
That will leave you with the sequences which meet the trend signature
criteria AND which are above the lower intensity cutoff in EVERY experiment.
This sounds complicated, but I ran through it for a 5 experiment trend in
about a minute.
Thanks for the quick answer!
It does work the way your colleague described but I want to select genes even if there is only signal in one experiment (a gene can have no expression at one timepoint and a significant expression at another timepoint). But I think I can make a bioset for this (I'll try).
And also the way of selecting on intensity is a bit subjective. Each array is different and when I select on all spots that have an intensity > 0.2 (for instance), this can include more spots in one array in comparison with another.
However, I will work with this option for now, but it would be nice if in the future a more general algoritm is available to select for spots that have no signal.