
                               EMBOSS: water
     _________________________________________________________________
   
                                 Program water
                                       
Function

   Smith-Waterman local alignment
   
Description

   water uses the Smith-Waterman algorithm (modified for speed
   enhancments) to calculate the local alignment.
   
   The Smith-Waterman algorithm is a member of the class of algorithms
   that can calculate the best score and local alignment in the order of
   mn steps, (where 'n' and 'm' are the lengths of the two sequences).
   These dynamic programming algorithms were first developed for protein
   sequence comparison by Smith and Waterman, though similar methods were
   independently devised during the late 1960's and early 1970's for use
   in the fields of speech processing and computer science.
   
   A local alignment searches for regions of local similarity between two
   sequences and need not include the entire length of the sequences.
   Local alignment methods are very useful for scanning databases or
   other circumsatnces when you wish to find matches between small
   regions of sequences, for example between protein domains.
   
   Dynamic programming methods ensure the optimal local alignment by
   exploring all possible alignments and choosing the best. It does this
   by reading in a scoring matrix that contains values for every possible
   residue or nucleotide match. water finds an alignment with the maximum
   possible score where the score of an alignment is equal to the sum of
   the matches taken from the scoring matrix.
   
   An important problem is the treatment of gaps, i.e., spaces inserted
   to optimise the alignment score. A penalty is subtracted from the
   score for each gap opened (the 'gap open' penalty) and a penalty is
   subtracted from the score for the total number of gap spaces
   multiplied by a cost (the 'gap extension' penalty).
   
   Typically, the cost of extending a gap is set to be 5-10 times lower
   than the cost for opening a gap.
   
Usage

   Here is a sample session with water.

% water sw:hba_human sw:hbb_human
Gap opening penalty [10.0]:
Gap extension penalty [0.5]:
Output file [hba_human.water]:

Command line arguments

   Mandatory qualifiers:
  [-sequencea]         sequence   Sequence USA
  [-seqall]            seqall     Sequence database USA
   -gapopen            float      The gap open penalty is the score taken away
                                  when a gap is created. The best value
                                  depends on the choice of comparison matrix.
                                  The default value assumes you are using the
                                  EBLOSUM62 matrix for protein sequences, and
                                  the EDNAFULL matrix for nucleotide
                                  sequences.
   -gapextend          float      The gap extension, penalty is added to the
                                  standard gap penalty for each base or
                                  residue in the gap. This is how long gaps
                                  are penalized. Usually you will expect a few
                                  long gaps rather than many short gaps, so
                                  the gap extension penalty should be lower
                                  than the gap penalty. An exception is where
                                  one or both sequences are single reads with
                                  possible sequencing errors in which case you
                                  would expect many single base gaps. You can
                                  get this result by setting the gap open
                                  penalty to zero (or very low) and using the
                                  gap extension penalty to control gap
                                  scoring.
  [-outfile]           align      (no help text) align value

   Optional qualifiers:
   -datafile           matrixf    Matrix file

   Advanced qualifiers:
   -[no]similarity     bool       Display percent identity and similarity

   General qualifiers:
  -help                bool       report command line options. More
                                  information on associated and general
                                  qualifiers can be found with -help -verbose
   

   Mandatory qualifiers Allowed values Default
   [-sequencea]
   (Parameter 1) Sequence USA Readable sequence Required
   [-seqall]
   (Parameter 2) Sequence database USA Readable sequence(s) Required
   -gapopen The gap open penalty is the score taken away when a gap is
   created. The best value depends on the choice of comparison matrix.
   The default value assumes you are using the EBLOSUM62 matrix for
   protein sequences, and the EDNAFULL matrix for nucleotide sequences.
   Number from 1.000 to 100.000 10.0 for any sequence
   -gapextend The gap extension, penalty is added to the standard gap
   penalty for each base or residue in the gap. This is how long gaps are
   penalized. Usually you will expect a few long gaps rather than many
   short gaps, so the gap extension penalty should be lower than the gap
   penalty. An exception is where one or both sequences are single reads
   with possible sequencing errors in which case you would expect many
   single base gaps. You can get this result by setting the gap open
   penalty to zero (or very low) and using the gap extension penalty to
   control gap scoring. Number from 0.100 to 10.000 0.5 for any sequence
   [-outfile]
   (Parameter 3) (no help text) align value Alignment file
   Optional qualifiers Allowed values Default
   -datafile Matrix file Comparison matrix file in EMBOSS data path
   EBLOSUM62 for protein
   EDNAFULL for DNA
   Advanced qualifiers Allowed values Default
   -[no]similarity Display percent identity and similarity Yes/No Yes
   
Input file format

   Any two sequence USAs of the same type (DNA or protein).
   
Output file format

   Here is the output file from the example run:
     _________________________________________________________________
   
Local: HBA_HUMAN vs HBB_HUMAN
Score: 293.50

HBA_HUMAN       2        LSPADKTNVKAAWGKVGAHAGEYGAEALERMFLSFPTTKTYFPHF 46
                         |:| :|: | | ||||  :  | | ||| |: : :| |: :|  |
HBB_HUMAN       3        LTPEEKSAVTALWGKV..NVDEVGGEALGRLLVVYPWTQRFFESF 45

HBA_HUMAN       47       .DLS.....HGSAQVKGHGKKVADALTNAVAHVDDMPNALSALSD 85
                          |||      |: :|| |||||  | :: :||:|::    : ||:
HBB_HUMAN       46       GDLSTPDAVMGNPKVKAHGKKVLGAFSDGLAHLDNLKGTFATLSE 90

HBA_HUMAN       86       LHAHKLRVDPVNFKLLSHCLLVTLAAHLPAEFTPAVHASLDKFLA 130
                         ||  || ||| ||:|| : |:  || |   |||| | |:  | :|
HBB_HUMAN       91       LHCDKLHVDPENFRLLGNVLVCVLAHHFGKEFTPPVQAAYQKVVA 135

HBA_HUMAN       131      SVSTVLTSKY                                    140
                          |:  |  ||
HBB_HUMAN       136      GVANALAHKY                                    145

%id = 45.99                     %similarity = 64.23
Overall %id = 43.15             Overall %similarity = 60.27
     _________________________________________________________________
   
   The %id is the percentage of identical matches between the two
   sequences over the reported aligned region.
   
   The %similarity is the percentage of matches between the two sequences
   over the reported aligned region where the scoring matrix value is
   greater or equal to 0.0.
   
   The Overall %id and Overall %similarity are calculated in a similar
   manner for the number of matches over the length of the longest of the
   two sequences.
   
Data files

   For protein sequences EBLOSUM62 is used for the substitution matrix.
   For nucleotide sequence, EDNAFULL is used. Others can be specified.
   
   EMBOSS data files are distributed with the application and stored in
   the standard EMBOSS data directory, which is defined by EMBOSS
   environment variable EMBOSS_DATA.
   
   Users can provide their own data files in their own directories.
   Project specific files can be put in the current directory, or for
   tidier directory listings in a subdirectory called ".embossdata".
   Files for all EMBOSS runs can be put in the user's home directory, or
   again in a subdirectory called ".embossdata".
   
   The directories are searched in the following order:
     * . (your current directory)
     * .embossdata (under your current directory)
     * ~/ (your home directory)
     * ~/.embossdata
       
Notes

   water is a true implementation of the Smith-Waterman algorithm and so
   produces a full path matrix. It therefore cannot be used with genome
   sized sequences unless you have a lot of memory and a lot of time.
   
References

    1. Smith TF, Waterman MS (1981) J. Mol. Biol 147(1);195-7
       
Warnings

   Local alignment methods only report the best matching areas between
   two sequences - there may be a large number of alternative local
   alignments that do not score as highly. If two proteins share more
   than one common region, for example one has a single copy of a
   particular domain while the other has two copies, it may be possible
   to "miss" the second and subsequent alignments. You will be able to
   see this situation if you have done a dotplot and your local alignment
   does not show all the features you expected to see.
   
   water is for aligning the best matching subsequences of two sequences.
   It does not necessarily align whole sequences against each other; you
   should use needle if you wish to align closely related sequences along
   their whole lengths.
   
   A true Smith Waterman implementation like water needs memory
   proportional to the product of the sequence lengths. For two sequences
   of length 10,000,000 and 1,000 it therefore needs memory proportional
   to 10,000,000,000 characters. Two arrays of this size are produced,
   one of ints and one of floats so multiply that figure by 8 to get the
   memory usage in bytes. That doesn't include other overheads. Therefore
   only use water and needle for accurate alignment of reasonably short
   sequences.
   
   If you run out of memory, try using supermatcher or matcher.
   
Diagnostic Error Messages

Uncaught exception
 Assertion failed
 raised at ajmem.c:xxx

   Probably means you have run out of memory. Try using supermatcher or
   matcher if this happens.
   
Exit status

   0 if successful.
   
Known bugs

   None.
   
See also

   Program name Description
   matcher Finds the best local alignments between two sequences
   seqmatchall Does an all-against-all comparison of a set of sequences
   supermatcher Finds a match of a large sequence against one or more
   sequences
   wordmatch Finds all exact matches of a given size between 2 sequences
   
   matcher is a local alignment program that is less rigorous than
   
   water
   and therefore runs more quickly. It may be useful for database
   searching.
   
   supermatcher is designed for local alignments of very large sequences
   and is even less rigorous in its implementation.
   
   supermatcher Finds a match of a large sequence against one or more
   sequences matcher Finds the best local alignments between two
   sequences
   
Author(s)

   This application was written by Alan Bleasby (ableasby@hgmp.mrc.ac.uk)
   
History

Completed 7th July 1999.
Modified 27th July 1999 - tweaking scoring.
Modified 22 Oct 2000 - added ID and Similarity scores.

Target users

   This program is intended to be used by everyone and everything, from
   naive users to embedded scripts.
   
Comments
