Straight forward Smith waterman aligner?

[thh32]

So I am looking for a straight forward smith waterman program which hasnt been "enhanced" and just accepts fasta files and compares to find common parts. The reason for not wanted enhanced versions is that we want to compare a few different aligners and need the smith waterman program to use as the gold standard and enhancements may use heuristics and therefore skip a few alignments which will reduce the reliability in the program. Does anyone know of a program like this? One which provides e-values / bitscores would be perfect.

[dpryan]

How big is the reference genome you're using and how much RAM do you have? This actually becomes the limiting factor for any implementation.

One old school program that can do this is "swat" from the phred/phrap package, which I believe also produces E values.

[thh32]

We have access to 500Gb RAM and we are aligning nucleotide sequences to Swissprot. hmmm ok I will have a look at that. I also found Swipe which provides e values so I am giving it a try.

[GenoMax]

If you can do protein-protein searches then you may want to try SSearch from FASTA suite: http://fasta.bioch.virginia.edu/fast...sta_down.shtml

[dariober]

So I am looking for a straight forward smith waterman program which hasnt been "enhanced" and just accepts fasta files and compares to find common parts.

Recently I was facing a similar question. I needed to find the best alignment between query and reference, both small (100s bp) and not very many (~1000s), no euristics, no filtering, just give me the best match. I ended up writing a simple program in Java which is here https://github.com/dariober/SequenceMatcher. The alignment part is delegated to BioJava.

I'm not sure it suites you though since in the most permissive mode it finds the best match between each query and each reference sequence. if one query aligns to one reference sequence in multiple locations equally well, you will get only one of them.

[DanOc]

My solution is a bit more work perhaps, but how about pulling one of the versions of Smith-Waterman from SHRiMP? It has a basic Vector S-W with basic scoring that it uses to filter for a more detail Scalar S-W down the line. The Vector version could be what you're looking for, if I'm reading the request correctly.

As for the extra work, we used this program as a framework for testing our hardware Smith-Waterman implementation and it was fairly easy to use due to the nice, clean code. It should be possible for you to pull out the bits you need in the same manner.

http://compbio.cs.toronto.edu/shrimp/

I'm not sure what the update rate of the codebase is like now, but I've been in contact with one of the original developers who was very nice about helping out with our questions.