Thanks Ben, I think I ultimately got the full picture.

In fact, I was unsure about what residues were exactly used for the structural alignment. Putting things together and with your answer, I think now I understand : Only pairs of residues that aligns in the sequence alignment are used for the feature computations/comparisons. More exactly, amongst these pairs only those whose residues are not too far away (under the rms cutoff) are actually considered in the score computation. That being said, salign tries to modify the sequence alignment (hence, the pairs of residues that will be used in the score computation) in order to lower the structural alignment score.

Thus the resulting structural alignment somehow relies on a ("structurally optimized") sequence alignment.

--Ben.S

That Le 7 avr. 2011 à 21:42, Modeller Caretaker a écrit :

> On 4/7/11 8:26 AM, Benjamin SCHWARZ wrote: >> As a first (naive) question on that mailing list, I wonder if (and >> how) it is possible to retrieve residues that are structurally >> aligned as a result of a aln.salign() run (object aln being of type >> alignment). To precise my question, I know a (multiple) sequence >> alignment is produced as a result of this salign command and can be >> output with aln.write(), but I am more concerned by the possible >> mapping of residues in the structural alignment, that might not be >> apparent in the sequence. > > Modeller does not make use of such a mapping internally. If two residues are aligned, then SALIGN uses the Cartesian distance between those two residues as an alignment score (put simply). It will modify that alignment to optimize that score. There is no difference between two residues being aligned "sequentially" or "structurally". > > Ben Webb, Modeller Caretaker