poor superposition after modeller comparative

[Elaine Meng]

The attached session shows the result of these commands for modeling a human heterodimer from the mouse one:

open 3bp5
open uniprot:pdcd1_human
sequence associate /A pdcd1_human
open uniprot:q2lc89_human
sequence associate /B q2lc89_human:1
modeller comparative pdcd1_human:1,q2lc89_human:1 multichain true numModels 3

(incidentally it never works when I put modeller commands in a script, I have to run them manually, dunno why)

I get models #2.1-3 of which the first two are not well superimposed although the log right after modeling reports good superpositions.

If I then manually run

mm #2.1/a,b to #1/a,b pair ss
mm #2.2/a,b to #1/a,b pair ss
mm #2.3/a,b to #1/a,b pair ss

(it helps to hide all but the one you're matching so you can see the improvement) the matches of the first two #2.1 and #2.2 improve a lot, #2.3 stays the same. However, all have the SAME RMSDs as originally reported in the Log from the automatic superposition. So I can only guess that the third match somehow messes up the previous two??

[pett]

The non-working-in-scripts problem opened as separate ticket #2487.

Superposition problem was because the matchmaker call had the reference/match chains reversed, so the template model was moving instead of the comparative models (and therefore only the last comparative model looked well superimposed).