Possibly, this is what you want, although it's not clear where the values to go in B come from:
A = zeros(10);
A(4, 4) = -2;
A(6, 6) = 2;
[r, c] = find(A); %find original coordinates of non zero values
c = c + nonzeros(A); %shift by value at coordinates
B = zeros(size(A));
B(sub2ind(size(B), r, c)) = -nonzeros(A) %are the values just the negative of the original ones?
