In article snc at worak.kaist.ac.kr, yshan at sorak.kaist.ac.kr (Yoonsoo Hahn) writes:
> hum..I wonder whether i can understand myself in English.
>> I'm trying to find any correlations between DNA sequence properties.
> But, It's window-size sensitive, that is, if the window size is big
> the property becomes to the average value of whole sequence.
> If the window size is small the variance of the property is too big.
>[figures deleted]
>> I typically choose 100 bp or 500 bp or 1 kb for the window size.
> And now I want to know IF THERE IS ANY GENERAL SIZE OF THE WINDOW for
> this kind of analysis.
>> Thanks in advance.
I have met exactly the same type of problem, and I did not either find a
general size of window. I don't think one exists, as different properties
will not change along the same pattern. In the case of the parameters I
was interested in, it was between 100 and 200 bp.
The solution I used was to trace the following diagram :
+--------------------------------- position of the window center
||| colors indicating the value
| of the parameter (discrete
| classes)
||size of the window
The original idea wasn't mine, it came from two math PhD students who had
developped this software for a different purpose : G Didier and L Vuillon.
It worked on a color Macintosh.
The e-mail of G Didier is : didier at lmd.univ-mrs.fr
hoping this may help,
Marc Robinson
