[ccp4bb]: Behavior of refmac under different program versions

[Pinotsis <pinotsis@embl-hamburg.de>]

***  For details on how to be removed from this list visit the  ***
***          CCP4 home page http://www.ccp4.ac.uk         ***

Dear all,

I was using Refmac_5.0.36 on an alphserver including TLS refinemnet.
A typical cycle of refinement was like this:
-----------------------------------------------------------------------------
Overall                 : scale = 0.705, B = -0.594
Babinet"s bulk solvent  : scale = 0.316, B = 200.000
Partial structure 1     : scale = 0.710, B = 15.788
Overallanisotropic scale factors
   B11 = -3.71 B22 = 9.88 B33 = -6.17 B12 = 0.00 B13 = 0.00 B23 = 0.00

Overall sigmaA parameters : sigmaA0 = 0.903, B_sigmaA = 4.751
Babinet"s scale for sigmaA : scale = -0.001, B = 150.000
SigmaA for partial structure 1: scale = 0.150, B = 0.001
-----------------------------------------------------------------------------
-----------------------------------------------------------------------------
Overall R factor           = 0.2633
Free R factor              = 0.3039
Overall figure of merit    = 0.6529
----------------------------------------------------------------

Then we decided to upgrade to Refmac_5.1.19
The same input files, the same alphaserver gave:

 Problem in MAKE_U_POSITIVE -1.4325634E-02
-----------------------------------------------------------------------------
Overall               : scale = 2.000, B = 0.000
Babinet"s bulk solvent: scale = 0.881,B= 50.000
Partial structure 1: scale = 0.350, B = 70.000
Overallanisotropic scale factors
   B11 = -3.73 B22 = 10.03 B33 = -6.30 B12 = 0.00 B13 = 0.00 B23 = 0.00
-----------------------------------------------------------------------------


-----------------------------------------------------------------------------
Overall R factor        = 0.2975
Free R factor           = 0.3439
Overall figure of merit = 0.5353
-----------------------------------------------------------------------------

(and the Rms Delta for distances and angles slighty higher)

And finally at the second case the final statics went even higher.

I tried with the same input files then on an SGI where Refmac_5.1.27 is intalled.
Here the first 2 cycles of TLS had slighty lower R's and FOM(!) but the
TLS values where different. At the 3rd TLS cycle started problems
messages:
 Problem in MAKE_U_POSITIVE -2.6608959E-02
Problem with atom in ELDEN 3140 (several times) and
Problem with atom in GRAD  3112 etc.


I would appreciate any help or hints.
Moreover, if this is a compilation problem, is it possible and other
CCP4.2 programs to produce such problems? (for example arp/warp interacts
with refmac).

best regards
Nikos Pinotsis
EMBL-Hamburg