How do I generate a global or local alignment with the Needleman-Wunsch- or the Smith-Waterman-algorithm?

A common approach to perform pairwise sequence alignments is to use the deterministic algorithms of Needleman-Wunsch and Smith-Waterman. Needleman and Wunsch defined the problem for global alignments, whereas Smith and Waterman developed an algorithm for local alignments, which was the basis for the BLAST program. The example below shows how to use the implementation of both algorithms from the alignment package. It requires BioJava 1.5.

The idea of these deterministic approaches is to maintain a matrix representation of an edit graph, which covers the operations insert, delete, replace and gap extension (insert and delete are gap openings in the query or the target sequence, respectively). By dynamic programing the matrix elements are computed, which are costs or scores of the respective operation. The highest scoring (lowest cost) path through the matrix gives the best alignment, which can be obtained by tracing back the matrix elements.

Alignments with different scores/expenses for gap opening and gap extension (affine gap penalties) consume significantly more time and memory as with equal scores (costs). The reason for this is that instead of one matrix three matrices are needed to remember the best path through the edit graph. Two matrices to remember if a gap has just been started or has been extended in the query or in the target sequence, respectively, and a third matrix to remember the optimum of these two and match/replace operations. All these matrices have the dimension query.length() times target.length().

These implementations of the algorithms need to be initialized with expenses (costs, penalties) for every edit operation. However, the substitution matrices use scores, which are the opposit of expenses. That means one can optain expenses from scores by multiplying them with -1. Using costs/penalties instead of scores allows to compute the edit distance for global alignments. This distance doesn’t make sense for local alignments, because in extreme cases a local alignment is an alignment of one symbol from each sequence and the distance would be zero, even if the sequences are much longer. This is why local alignments are based on scores. Nevertheless the constructor of SmithWaterman needs penalties and not scores.

The super class SequenceAlignment of the specific alignment algorithms already provides a method to format the alignment output. So if you want to write your own alignment algorithm or to use the HMM-alignment algorithm, you can derive your class from this super class and apply the given method.

Many substitution matrices have been released for the purpose to evaluate the transition from one symbol to another one. These can be downloaded at NCBI and are needed for the following example. Alternatively, the SubstitutionMatrix class provides a constructor to create your own substitution matrix with equal scores for every match and every replace.

A demo of local and global alignments

```java import;

import; import; import; import; import; import; import; import;


* Created on Mar 28, 2006

/** This is a demo class that performes both a local and a global alignment

 * from two given sequences. The result is printed on the screen. 
 * Therfore a substitution matrix file is required, which can be downloaded
 * at {@link}.
 * This demo only works for DNA sequences. However, the alignment algorithms
 * are able to use any kind of alphabet as long as there is a substitution
 * matrix available. For this example the matrix NUC.4.4 is the best one.
 * @author Andreas Dräger

public class DeterministicAlignmentDemo {

 /** This performs an alignment of two given sequences and 
   * prints the result on the screen.
   * @param args: a query and a target sequence 
   *   and one file containing the substitution matrix to be used.
   * {@link}
 public static void main (String args[]) {
   if (args.length < 3)
     throw new Error("Usage: DeterministicAlignmentDemo " +
                     "querySeq targetSeq substitutionMatrixFile");
   try {
     // The alphabet of the sequences. For this example DNA is choosen.
     FiniteAlphabet alphabet = (FiniteAlphabet) AlphabetManager.alphabetForName("DNA");
     // Read the substitution matrix file. 
     // For this example the matrix NUC.4.4 is good.
     SubstitutionMatrix matrix = new SubstitutionMatrix(alphabet, new File(args[2]));
     // Define the default costs for sequence manipulation for the global alignment.
     SequenceAlignment aligner = new NeedlemanWunsch( 
       (short) 0,  // match
       (short) 3,  // replace
       (short) 2,      // insert
       (short) 2,  // delete
       (short) 1,      // gapExtend
       matrix  // SubstitutionMatrix
     Sequence query  = DNATools.createDNASequence(args[0], "query");
     Sequence target = DNATools.createDNASequence(args[1], "target");
     // Perform an alignment and save the results.
       query, // first sequence
       target // second one
     // Print the alignment to the screen
     System.out.println("Global alignment with Needleman-Wunsch:\n" + aligner.getAlignmentString());     
     // Perform a local alginment from the sequences with Smith-Waterman. 
     // Firstly, define the expenses (penalties) for every single operation.
     aligner = new SmithWaterman(
       (short) -1,     // match
       (short) 3,      // replace 
       (short) 2,      // insert
       (short) 2,      // delete
       (short) 1,      // gapExtend
       matrix  // SubstitutionMatrix
     // Perform the local alignment.
     aligner.pairwiseAlignment(query, target); 
     System.out.println("\nlocal alignment with SmithWaterman:\n" + aligner.getAlignmentString());
   } catch (Exception exc) {

} ```