[phenixbb] refine : peculiar TLS tensors

Pavel Afonine pafonine at lbl.gov
Wed Oct 20 12:08:52 PDT 2010

  Hi Bryan,

> let me try a simpler question : do the TLS results need to be
> transformed or scaled somehow so they are on the same basis as the
> ones from refmac/tlsanl/anisoanl (i know this is partially a ccp4
> question)?

I wrote most (if not all) TLS related code in phenix. Unfortunately, I 
can't answer your questions because I do not know which transformations 
the above programs do, which units, origins and scales they use, and 
even what they do (speaking of the last two). So, I afraid I'm useless 

I'm sure there must be a thing or a number of things somewhere like a 
scale factor such as units for L (degrees**2 or radians**2), or 
8*pi**2*something in U conversions, or different origin, etc ... I 
recall I had to figure out myriads of things like this before TLS 
started working in phenix.refine. Typically these "obvious" details are 
never mentioned in general papers as probably "too technical".

This is why it is rarely not tricky to jump between the programs.
It's probably like if you take two different cars (different makes and 
models) that both function well and sound, and then swap their engines 
without thinking too much, then the chances are that both will stop working.

Let do something practical to address this... What I can do for you is: 
if you clearly explain me what exactly you are trying to achieve then I 
just simply add the corresponding functionality to PHENIX and then you 
(and potentially other PHENIX users) will be able to do what you want 
within one self-consistent framework, and avoiding jumping between 
units, conventions and other hidden meanings. Of course this might take 
some time - from a couple of days to a week, so you would need to wait a 
little bit, but then you will have a tool in hands to use any time you 
want. So, let me know.

> p.s: new subscription address : slow to appear on bulletin board...?

Not sure what you mean ?


More information about the phenixbb mailing list