[BioC] Macroarray Image Analysis (summary)

[Nicholas Lewin-Koh]

hi everyone,
A few weeks ago I sent a message to the list regarding image analysis
tools for extracting spots from 
macro arrays (spotted Nylon membranes). My experiences so far with
macroarrays has not been great,
1) I haven't found a good image analysis program for extracting the spot
intensities reliably (I will summarize 
this below) and 2) I am not making many biologist friends when I tell
them nothing is significant ;^)

The programs I evaluated were:
SpotFinder: a Java plugin for Jimage at
http://www.math.usu.edu/~adele/spotfinder/index.htm
Spot: The program advocated by the Speed clique
Bzscan:  A stand alone java program
http://tagc.univ-mrs.fr/bioinformatics/bzscan/

Well Spot failed, I think primarialy because it didn't handle
overlapping spots. The spots do difuse
and one can get spot overlap. Also Spot is expensive, for the other two
the price is right.
Spot finder seems to be doing ok. The program could use a lot of
improvement, the automatic 
gridding did not work well on my data, and really couldn't handle shear
in the images. 
But the spot finding algorithm is nice and is based on a spot shape
model. It handles diffusion of
the spots well. I am going to try and get a student intern to work on it
and improve the speed 
and make the algoritms more robust to errors on the image. If my company
lets me, I will release
any improvements. Spot finder was hard to get to work "out of the
box" I had to change some of the source. I played with BZscan a bit. It
was buggy and the interface
clumsy (for me). Also when I was having problems I got no reply form the
authors and some of the download links
are broken on the web page. It is not clear if they are still supporting
it. I also had some problems
with spotfinder and getting a reply from the author, but to be fair
Adele Cutler is on vacation
until the 26th of August and I probably wouldn't answer my email either.
She did at least set her auto-reply.

I did notice in the archives that Anthony Rossini was looking into
making an R-package for Jimage. If he 
is still planning that it makes SpotFinder even more appealing because
it can be run from R with not
too much modification. Another note is that Claus Ekstrom has a nice
paper on spotshape modelling 
for microarrays. The work is derived from other work by Mats Rudemo and
Erik Gustafson on modelling
spot shapes on protien gels, which are probably similar to macroarrays.
I think they are working on an R package
as well which may be useful. Lastly I was reccomended to try Arraypro
analyser by Peter
Wilkinson
(https://stat.ethz.ch/pipermail/bioconductor/2004-July/005533.html)
I haven't downloaded the demo yet and tried, sorry.

Hope this helps.

Nicholas