Biomedical Engineering Reference
In-Depth Information
Endnote
The pattern-matching approaches using scores for gaps and inexact matching or black-box neural
network technology discussed here are statistically valid for assessing the degree of string similarity.
However, in selecting gaps and other methods to make the matches "look good," it's important to
remember that these techniques don't necessarily relate to the biology of the nucleotide or
polypeptide chains represented by the symbols manipulated by BLAST or other algorithms. It's easy
to rationalize the need for gaps because of the computational infeasibility of solving long string
comparisons without the provision for gaps. However, even a short gap in a polypeptide sequence
can disrupt the secondary and tertiary structures of a protein, and probably alter its function as well.
Heuristic approaches, such as match matrices, attempt to add some sense of biological relevance to
the mathematical equations that define the relative similarity of nucleotide and polypeptide
sequences. It's up to individual researchers to consider the biological implications of the techniques
and assumptions they make in simply filling out a form on a Web page during the course of their
daily work.
