Talk:Smith–Waterman algorithm

WPMED

I am not sure this article falls within the scope of WP:MED. Anyone who agrees, please delete the WPMED tag on this page. --Una Smith (talk) 04:29, 29 December 2007 (UTC) Biology - Yes, Medicine - well, there is very remote connection - but no. Crenshaw (talk) 02:03, 19 January 2010 (UTC)[reply]

Pseudocode of Algorithm and Remarks to Code are missing

There is the simple but efficient pseudocode of the algorithm missing. It is tailored to fit to the properties of genes in common genomes. Maybe someone can link the properties of common (for example eucaryotic cell) genes having introns and exons to the properties of Smith-Waterman or pairwise Smith-Waterman pSW --84.157.243.81 (talk) 09:32, 24 July 2008 (UTC)[reply]

Eq. -- why the 0?

Shouldn't $H(i,j)=\max {\begin{Bmatrix}0&\\H(i-1,j-1)+\ w(a_{i},b_{j})&{\text{Match/Mismatch}}\\H(i-1,j)+\ w(a_{i},-)&{\text{Deletion}}\\H(i,j-1)+\ w(-,b_{j})&{\text{Insertion}}\end{Bmatrix}},\;1\leq i\leq m,1\leq j\leq n$ be $H(i,j)=\max {\begin{Bmatrix}H(i-1,j-1)+\ w(a_{i},b_{j})&{\text{Match/Mismatch}}\\H(i-1,j)+\ w(a_{i},-)&{\text{Deletion}}\\H(i,j-1)+\ w(-,b_{j})&{\text{Insertion}}\end{Bmatrix}},\;1\leq i\leq m,1\leq j\leq n$

What's that zero doing in the B-matrix? —Preceding unsigned comment added by 129.170.215.143 (talk) 23:38, 18 June 2009 (UTC)[reply]

Smith-Waterman is for local alignment, which means that if H(i, j) is 0, then it's best for the alignment to start there. I think the example is really confusing, because it doesn't illustrate this fact, so you get an optimum local alignment that is also a global alignment. 69.218.212.242 (talk) 17:20, 8 August 2009 (UTC)[reply]

Performance reports

I find this article similar to post with "me faster" signs. As someone who actually takes part in "competition" of speeding up Smith-Waterman for different architectures I think it's a bit NPOV. (disclaimer: i haven't add mine implementation yet to the article but I intend to do so).

Performance results should be reported not as a single value but a graph with performance plotted against query size and database used. For example Fastflow implementation achieves 35GCUPS for __very__ long queries that are barely if ever used in practice. I think it's a bit misleading.

I guess that I'll rewrite this article and put pefrormance graph for different algorithms. But it would be nice if someone reviewed the article afterwards (coz I'll add my own work on SW too) to eliminate potential NPOV.

Crenshaw (talk) —Preceding undated comment added 22:20, 30 January 2010 (UTC).[reply]