I've no real disagreements with Tim.
On Sat, 14 May 2011, Tim Stevens wrote:
>> It does work in ARIA 2.3 with analysis 2.1.5 and you can control what goes
>> back into the project to some extent. My personal recommendation is to
>> export restraints and violations from the final iteration, but not peak
>> lists (can mangle your assignments if you're not careful)
> My preferred way of working is also to not use ARIA assigned peak lists, but
> rather to add to assignments as the structure improves using the NOE
> Contributions system, seeding assignments for the solid parts and letting
> ARIA figure out the rest.
> However, I wish to make it crystal clear that the ARIA export to CCPN does
> not overwrite or change any existing peak lists or resonance assignments.
> Brian, can you please explain a bit more about what you mean by the "mangle"
> comment. Is there something that needs changing?
> I do know that the peak lists that come back might have strangely paired
> protons and hetero atoms, but the extent of the issue is in the new peak list
> and is easily deleted.
As Tim says, the reimported peak lists are easily deleted restoring
the project to its unsullied state, but unless I'm out of date on this,
the reimported peak lists contribute equally to the shift lists with the
originals and so shift values can be pulled away from their correct
values. If I'm still right about this then I guess my suggestion would be
that the reimported peak lists have weight 0 for shiftList calculation. I
guess that apart from user confusion, the strange pairings you mention can
also cause havoc with e.g. quality reports - maybe the reimported peak
lists should be flagged to be ignored by them?
>> or structures (don't have fine enough control about how many go back and
>> from which iteration leading to project bloat).
> To be clear, you can use ARIA to export the last iteration and water
> refinement structure ensembles. You can also delete old structures. This
> keeps the bloat to a minimum.
...but if I remember right you get all the structures calculated in those
iterations rather than the "best" N whatever N is. If you're running ARIA
with the defaults (I would not advise this as a rule!) you get 20
structures per iteration which is manageable, but if you're running a more
realistic number then bloat is a real issue. If I'm wrong or out of date
about this I apologise.
>> The slickest way to get structures in again is to make the ones you want
>> (lowest energy selection from final iteration or refinement) into a
>> multimodel pdb file (I find the old aqua joinpdb file script handy for
>> this) and import through the structures popup in analysis.
> You can select and load multiple PDB files in analysis and [Merge Into
> Ensemble] rather than use external scripts.
OK that's a good feature I'd so far missed. Any chance of the import being
able to interpret ARIA .float files to get prochirals the right way round,
or do you see that as ARIA's responsibility?
> Also, I would say that subjectively selecting PDB files
Whoaa! who said anything about subjectively? I would absloutely advise
objective selection - lowest restraint energy, lowest overall energy, or
some other appropriate objective criterion. Definitely not "do I like the
look of it"! And I would advise including a large enough number of
structures to get good stats on NOE violations etc.
> from an ensemble is not for the uninitiated. Certainly structure
> calculations can go awry, to the point where they are not so useful, but
> also this can be minimised by gentler MD annealing.
Certainly - the default ARIA number of steps are only suitable for quite
small proteins and/or very clean input data.
> I would say that pruning an ensemble can be dangerous in early
> calculations because of the risk of biasing toward one conformation. And
> in general I think it is better to focus on getting the interpretation
> of the NMR data correct, and let the ensemble reflect the precision in
> the data.
> Personally I like to see some outliers because it gives me more information
> about potential problems in the assignment.
Absolutley with you there. One comment on the link between the
violations/peak/assignment popups and the integrated structure viewer - it
would help to keep users away from focusing their violation analysis on a
single conformation (and risk biasing it towards that structure) if the
"show on structure" functionallity worked nore slickly as you switch the
structure displayed from an ensemble. But that would be icing on the
already good cake.
Dr. Brian O. Smith ------------------------ Brian Smith at glasgow ac uk
School of Life Sciences, College of Medical, Vetinerary & Life Sciences,
Joseph Black Building, University of Glasgow, Glasgow G12 8QQ, UK.
Tel: 0141 330 5167/6459/3089 Fax: 0141 330 4600
The University of Glasgow, charity number SC004401