scoring matrix

[huda]

hi,
how design scoring matrix.

i'm working seq. alignment with non biological seq.

if i have 11000 states, so i have to use 11000 unique codes.
if so, must scoring matrix e 11000*11000?

thanks in advance

[mfursov]

A straightforward way could be to merge you matrix into smaller one using approximation that some symbols from your alphabet are equal

If the result length is small, for example 10-100-200 bytes, you can process the result again and here use global precise matrix. In this case the needed linear matrix size will not exceed the result length.

[huda]

thanks,
please, i'm not fully understand what you mean, give me more details.
what I mean that in case of biological sequence , the substitution matrix (scoring matrix)
be 20*20 if the sequences are amino acids, and 4*4 in case of nucleotides.
But, in my case ,the number of unique codes are 11000.

I sent that thread since time , so I read and read and I found the unitary matrix is very suitable with my case.i.e match=1, mismatch=0 gap=-1.
I hope to know your comments

[mfursov]

>please, i'm not fully understand what you mean, give me more details.

Lets say that you have 1000 letters in your alphabet and the scoring matrix is 1000x1000
in this case you can split letters to N groups, for example N=10 and get a scoring matrix 10x10 In this matrix in every cell you can use the best score possible from the group.

Once the result is found you can build new precise matrix only with letters that are in result and recalculate result score.