In reply to Sean Eddy's posting: there is a set of proteins whose sequences
are rather different (I dare not write "remotely similar", which looks
rather strange...) but which can be aligned on the basis of 3-D
structures in the family. It is the aminoacyl-tRNA synthetases.
1) an alignment of the 10 class I aaRS from E. coli, based on three
crystal structures, should be published soon (I hope) by Landes et al.
in J. Mol. Biol. (if God permits. The paper is submitted).
2) as expected, NO pairwise alignment performed with GAP or BESTFIT is
totally consistent with the "3-D alignment".
3) we have made a thorough study of the ability of different scoring
indices to produce a correct *classification* (not phylogeny) of the
globins and the synthetases. As expected (again) the BESTFIT and FASTA
scores failed to classify correctly the vertebrate globins and the aaRS.
The classifications are much more satisfactory when the Z-scores are
taken as distance indices. This will appear soon in NAR.
4) the conclusion is clear: alignment scores can in no way be used to
compare different alignments, except when the sequences are very close
to one another. This is not a great surprise.
5) may I humbly direct you to a paper that I published some time ago (J.
Mol. Biol. (1988) 204, 1019-1029) where I tried to build a scoring
matrix based on 3-D superpositions ?
To come back to the SCIENCE paper: does all this stuff mean that papers
which need experimental details should not be published in Science ?
************************************************************************
J-L. Risler Centre de Genetique Moleculaire risler at frcgm51.bitnet
Gif sur Yvette France bionet at frcgm51.bitnet
