ss_lg_align, a FORTRAN90 code which implements some of the string matching algorithms described by Chao.

These algorithms look for optimal local alignments of two strings using linear space. (Compare the global alignment routines in SS_GG_ALIGN.)

It's important to be able to compute alignments using "linear space", that is, just a few vectors whose length N is equal to that of a typical string. A quadratic algorithm would require a two dimensional array of total dimension N*N. Realistic alignment problems can involve strings of N=100,000 elements or more, so a quadratic algorithm would be expensive or impossible to use.

The "matching" being considered does not actually require that every element of string A match an identical element of string B. Instead, the matching algorithm is essentially looking for the highest scoring way of associating a portion of one string with a portion of the other, allowing the operations of "mutation" (change one letter to another), "deletion" (drop a string of consecutive letters) and "insertion (insert a string of consecutive letters). Thus, at the same time, we are measuring a sort of evolutionary distance between two strings and a formal "editing distance" between them.

This set of routines assumes that an insertion or deletion of length K is penalized using an "affine gap penalty formula" of the form:

Penalty = Gap_Open + K * Gap_Extend
This choice of penalty function has a major effect on the form of the matching algorithms, particularly in the linear space case.

The score for the actual best matching is determined without explicitly constructing the best matching. It is a matter of some difficulty to recover the matching corresponding to the best score. This is particularly true if the algorithm is a linear space one, which discards a great deal of intermediate information. However, it is possible to set up a recursive algorithm which determines the best alignment, using only linear space.

Routines that use quadratic space are included as well, so the algorithms can be compared for storage, speed, and correctness.

The names of the scoring and path routines include information about whether they use a forward, backward, or recursive algorithm, whether they compute the score or the path, and whether they use linear or quadratic space. Thus, the routine SS_LG_FSQ uses the forward algorithm to compute the score, with quadratic space requirements.


The computer code and data files described and made available on this web page are distributed under the GNU LGPL license.


ss_lg_align is available in a FORTRAN90 version.

Related Data and Programs:


SS_GD_ALIGN, a FORTRAN90 library which globally aligns two sequences using a distance matrix.

SS_GG_ALIGN, a FORTRAN90 library which globally aligns two sequences using an affine gap penalty

SS_QG_ALIGN, a FORTRAN90 library which quasi-globally aligns two sequences using an affine gap penalty.


  1. Kun-Mao Chao, Ross Hardison, Webb Miller,
    Recent Developments in Linear-Space Alignment Methods: A Survey,
    Journal of Computational Biology,
    Volume 1, Number 4, 1994, pages 271-291.
  2. Eugene Myers, Webb Miller,
    Optimal Alignments in Linear Space,
    Volume 4, number 1, 1988, pages 11-17.
  3. Michael Waterman,
    Introduction to Computational Biology,
    Chapman and Hall, 1995,
    ISBN: 0412993910,
    LC: QH438.4.M33.W38.

Source Code:

Last revised on 02 April 2023.