long running time

Question

huda nawaf 2012-4-17

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/35790-long-running-time

在 MATLAB Online 中打开

hi,

i have code (smith_waterman) algorithm of sequence alignment.

I have to align one sequence to large set of squences (database) when run it, it take very long time .

can any one reduce the running time by reducing steps?

Note: I do not know why be the running time longer when used large database. i.e when I align one sequence to 100 sequences from array with size(7000*1000) for example the time will be longer when align one sequence to 100 sequences from array with size(200*1000)

this is code:

f1=fopen('d:\matlab11\r2011a\bin\webscopdata\webscop\action.txt');
c=textscan(f1,'%d%d');fclose(f1);c1=double(c{1});c2={2};
[p,o]=hist(c1,unique(c1));celcod=dlmread('d:\matlab11\r2011a\bin\webscopdata\webscop\webfinal.txt');
gap = -1;
for b=1
      f=0;cc=1;
      for b1=1:10
          if b~=b1
match=2;mismatch=-1;align1=[];align2=[];x=celcod(b,:);   y=celcod(b1,:);
x1(1:length(y),1:length(x))=0; trac(1:length(y),1:length(x))=0;
txt(1:length(y),1:length(x))=' ';       mat=struct('scor',x1,'pointer',txt);
                  for j=1:length(x)
     mat(1,j).scor=x1(1,j);     mat(1,j).pointer='none';
                  end
                  for i=1:length(y)
     mat(i,1).scor=x1(i,1);   mat(i,1).pointer='none';
                  end
max_i=1;max_j=1;max_scor=1;score=0;
                   for i=2:length(y)
                       for j=2:length(x)
                 letter1=x(j);letter2=y(i);
                           if strcmp(letter1,letter2)
                              dig_scor=mat(i-1,j-1).scor+match;
                           else
                              dig_scor=mat(i-1,j-1).scor+mismatch; 
                           end
   up_scor=mat(i-1,j).scor+gap;  eft_scor=mat(i,j-1).scor+gap;
        if (dig_scor<=0 && up_scor<=0 && left_scor<=0)
         mat(i,j).scor = 0;   mat(i,j).pointer='none';
                                continue
                           end
                          if dig_scor>=up_scor
                                  if dig_scor>=left_scor
    mat(i,j).scor=dig_scor; mat(i,j).pointer='diagonal';
                                       else
   mat(i,j).scor=left_scor; mat(i,j).pointer='left';
                                  end
                          else
                                  if up_scor>=left_scor
     mat(i,j).scor=up_scor;  mat(i,j).pointer='up';
                                  else
     mat(i,j).scor=left_scor;  mat(i,j).pointer='left';
                                  end
                          end
                          if mat(i,j).scor > max_scor
        max_i=i;  max_j=j; max_scor=mat(i,j).scor;
                          end;end;end
               max_scor;j=max_j; i=max_i;
       while i>1 & j>1
                  if strcmp(mat(i,j).pointer,'none')==1
                     break;  end
                 if strcmp(mat(i,j).scor,'0')==1
                     break;
                  end
                 if strcmp(mat(i,j).pointer,'diagonal')==1
    align1= [align1 x(j)];align2=[align2 y(i)]; trac(i,j)=1;
                      if strcmp(x(j),y(i))==1
                          score=score+2;
                      else
                          score=score-1;  end
                  i=i-1; j=j-1;
                  elseif strcmp(mat(i,j).pointer,'left')==1;
     align1= [align1 x(j)];align2=[align2 '-'];
                trac(i,j)=1;j=j-1;score=score-1;
                 else mat(i,j).pointer=='up'
   align1= [align1 '-' ]; align2=[align2 y(i)];trac(i,j)=1;
                          i=i-1; score=score-1;
                  end; end
                align1=fliplr(align1);align2=fliplr(align2);
                 [align1;align2];
                  for i=1:length(y)
                      for j=1:length(x)
        mat1(i,j)=mat(i,j).scor;  mat2{i,j}=mat(i,j).pointer;
                      end
                  end
                scor(b1)=max_scor; 
                        end
       end
end