Anthony Berno (aberno at genome.stanford.edu) wrote:
: it must have the following features:
: -It must not be instrument-dependent, as we are
: going to be using it for new instruments currently
: under development
This is fairly easy as long long your algorithm is flexible
enough to handle arbitrary sizes and resolutions of images,
such as 8-bit, 16-bit, etc.
: - It must be able to track an arbitrary number
: of lanes, and
This is also not too difficult once you find all the bases and
you figure out by some simple statistical approach.
: -Most importanly, it must be able to do lane tracking
: with NO human intervention, EVER, and either get it
: right, or determine by itself that it has made a
Well, this one is not as easy as it sounds if you want to be able
to handle the previous two. If you study enough number of DNA
sequencing gels, you'll see for a lot of them, you simply don't
have any information for the top and bottom. So how do you
know where to start and end? How the system know if it's making
a mistake, by magic? Just a simple case, when it's tracking a lane,
how do you know if it's still in the same lane or it already crossed
into another lane next to it?
I'm the author of DNAscan system, from Scanalytics/CSPI, the
algorithm I developed requires 4 point to locate the region of
the search area and it's been working very well, very robust.
Without the info, I'm sure I can get something for a set of
perfect images, but I'm sure it will fail sooner or later when
it sees real images.
: I've got a few leads already, but if anyone has
: anything to share, I'd like to hear from you.
Yecheng Wu, Ph.D.
P.S. Mike, it was nice to meet you at Waterville Valley, NH. Was a good