What is a good alignment? The amino acid sequence codes for the protein
three dimensional structure. Accordingly, when an alignment of two or
more sequences is made, the implication is that the equivalenced resdues
are preforming similar structural roles in the native folded protein.
The best judge of alignment accuracy is thus obtained by comparing
alignments resulting from sequence comparison with those derived from
protein three dimensional structures. There are now many families of
proteins for which two or more members have been determined to atomic
resolution by X-ray crystallography or NMR. Accurate alignment of these
proteins by consideration of their tertiary structures
[36][35][34] provides a set of test alignments against which to
compare sequence-only alignment methods. Care must be taken when
performing the comparison since within protein families, some regions
show greater similairty than others. For example, the core -
strands and
- helices are normally well conserved, but surface loops
vary in structure and alignments in these regions may be ambiguous, or
if the three-dimensional structures are very different in a region,
alignment may be meaningless. Accordingly, evaluation of alignment
accuracy is best concentrated on the core secondary structures of the
protein and other conserved features [37]; such regions may
automatically be identified by the algorithm of Russell and Barton
[36].