Reuben Leberman's models
Procedure
Dear Reuben,
I have first taken the structure file sent to me by Stephen,
extracted the sequence, and ran it against SwissProt using
PHD.
This resulted in a HSSP file.
I than used the WHAT IF to
automatically build the models for all known related molecules.
If you are interested, you can get some
information about
the methods used to build these models.
Models
view or get the original coordinates
view or get gvpk_halsa
view or get sys_coxbu
view or get sys_crigr
view or get sys_ecoli
view or get sys_haein
view or get sys_halma
view or get sys_human
view or get sys_mycge
view or get sys_theth
view or get sysc_yeast
Quality
To give an impression of the limits on the model quality I have produce the
structure quality report for the template coordinates.
The model quality reports (listed below) should be compared with this.
Before looking at the model quality, please keep in mind that:
- We maintain as much as possible the template backbone in the model. Even in case
of mutations from or to Gly or Pro. This sometimes leads to big local errors, but avoids
many global errors.
- The backbone quality normally gets worse upon modelling. The techniques to do this better simply
do not yet exist.
- We do not try to remove all small clashes. This sometimes leads to big local errors, but avoids
many global errors.
- For gvpk_halsa the sequence is incomplete (or the sequence identity too low). The model
is therefore VER incomplete, and probably totally rubish.
- Judging by the numbers only, most other models look rather good (for a model).
- During some manual WHAT IF 'jugling' I seem to have lost some of the C-terminal
oxygens. If that is a drama, tell me and I reconstruct them.
- B-factors of modelled atoms are set at 12.0. So the B-factor validation is useless.
- Other Xray tests are useless too (e.g. Mattwevs volume, cell dimensions...).
- The buried unsatisfied H-bond donors/acceptors seem to indicate some alignment problems.
You probably know much more about these molecules than I do, so could you look at those? If it
is a problem, I can do it, but for me it would be hours of work per model...
The reports: