distance metric

[huda]

hi,

 

If I have similarity matrix that I got it from sequence alignment, where each element of this matrix repersent the rate of similarity .i.e the max score represent mas similarity.

 

say n*n symmetric that show the rate of similarity among n objects .

 

 

ex.

x=[10 3 2

3 6 4

2 4 5];

 

here obj. 1 is most similar to obj. 2 because the max score except diagonal(10) is 3

 

and so on. the diagonal here is represent the comparison the object with itsel so has the max value

 

if I used distance metric such euclidean distance to cluster these objects, is it best to set the diagonal to zero, or leave it ?

 

 

by the way , which the best way to cluster the similarity matrix that result from sequence alignment?

[Phi for All]

!

Moderator Note

Please, just post your thread once in a single section. Multiple instances makes it very difficult to follow the discussion.

[huda]

!

Moderator Note

Please, just post your thread once in a single section. Multiple instances makes it very difficult to follow the discussion.

 

 

what I have to do now?