Hi Ian,

That is an interesting way of looking at it. Indeed another good reason to use rmsZ rather than the still much more popular rmsd.

Cheers,
Robbie



Sent from my Windows Phone

Van: Ian Tickle
Verzonden: 15-10-2015 21:03
Aan: [log in to unmask]
Onderwerp: Re: [ccp4bb] Remac5 refinement statistics

Hi Robbie

Yes nice result.  Notice that the Zangl values are greater than the Zdist ones.  This is entirely in line with expectation: the distance between the 1-3 atoms in an angle is obviously greater than the bond lengths.  This means that the 1-3 atoms are more resolved than the 1-2 ones and X-rays are better able to measure deviations from ideal angles, and so the rmsZ values are higher for angles.  This is one advantage of using rmsZ rather than rmsDev: this effect is much more obvious.

Cheers

-- Ian


On 15 October 2015 at 19:12, Robbie Joosten <[log in to unmask]> wrote:

Hi Tassos and Ian,

 

I was busy fending off a Scrin attack on Sydney, hence my late response.

 

At 2.5A (n=2972) Zbond is on average 0.51 (+-0.27) in the PDB and 0.47 (+-0.14) in PDB_REDO. Zangle is 0.75 (+-0.31) in the PDB and 0.70 (+-0.12) in PDB_REDO. At that resolution, the number of reflections per atom ranges from 1.2 to 18 with an average of 5.3 (+-1.5). The rmsZ values and the number of reflections per atom for this set are not correlated (at least not linearly).

 

Cheers,

Robbie

 

 

 

From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of Anastassis Perrakis
Sent: Thursday, October 15, 2015 19:39
To: [log in to unmask]
Subject: Re: [ccp4bb] Remac5 refinement statistics

 

Dear Ian,

 

I cant disagree with you - I get your point.

 

But I would be curious to see what the PDB_REDO Z scores would be (which also takes into account the Rfree for different geometry weight). I admit I aim for “less than 1” - sometimes 0.5 appears tight, but I am bit rusty in refinement lately, I have been sub-contracting a lot to a certain young gentleman ;-)

 

A.

 

 

On 15 Oct 2015, at 19:29, Ian Tickle <[log in to unmask]> wrote:

 

 

Hi Tassos

Sorry I have to disagree with you there.  Z scores of 1 @ 2.5 Ang would not be good: they should probably be less than 0.5.  By 'high resolution' the documentation really means high, i.e. atomic.  The X-ray data doesn't begin to tell you about the distances between bonded atoms until it can 'see' that the atoms are distinct, i.e. resolved.  At lower resolution the restraints take over, so the deviations from the restraints (i.e. the Z scores) become less.

Yippee, I beat Robbie too!

Cheers

-- Ian

 

On 15 October 2015 at 18:20, Anastassis Perrakis <[log in to unmask]> wrote:

Dear Ray -

 

On 15 Oct 2015, at 19:12, Ray Brown <[log in to unmask]> wrote:

 

I have a question about when to actually stop refinement with Refmac5. What is a properly carried out refinement with good statistics?

 

All I could find was this advice: 

 

'zBOND and zANGL are the Z-scores of the deviations of bond lengths and angles from ideal values: at high resolution these should approach 1.0 while at low resolution, where structures are tightly restrained to ideal values, they should approach 0.0.'

 

What values of zBOND and zANGL are expected for a well-built and correctly refined protein structure at a resolution of 2.5 angstroms?

 

Around 1 would be good...



 

Should the geometric weights in Refmac5 be adjusted/varied so as to produce statistics of rmsBOND 2.0, rmsANGL 2.0 and rmsCHIRAL 0.2? Are these values appropriate for this resolution?

 

I would suggest to let PDB_REDO optimise these for you - http://xtal.nki.nl/PDB_REDO/

 

Best -

 

Tassos (for once faster that Robbie!)



 

Some practical advice and your experience would very much be appreciated.

 

Ray Brown