wrong DP matrix init?

[sjanssen2]

I assume you intend to compute global pairwise alignments, i.e. Needleman Wunsch. Implicitly, you are initializing your first row and first column with 0, which I think is not correct. Traditionally, it is initialized as i * gapcost (c.f. http://rna.informatik.uni-freiburg.de/Teaching/index.jsp?toolName=Needleman-Wunsch#), in your code that should translate to V[i-1, 0] + A[i-1, 0]. Otherwise, leading gaps in both sequences do NOT get penalized and it would be half way towards an "end gap free" algorithm.

deepblast/deepblast/nw.py

Lines 89 to 93 in ab48e4f

	V = new(N + 1, M + 1).zero_() # N x M
	Q = new(N + 2, M + 2, 3).zero_() # N x M x S
	Q[N + 1, M + 1] = 1
	for i in range(1, N + 1):
	for j in range(1, M + 1):

What's missing for a real "end gap free" implementation is that you find your global optimum as the max in the last col and row of your DP matrix.
Therefore, I am confused about what version you really want to implement here?