You are here

GAB force fiield implementation seems broken

Forums: 

I am trying to run the most recent version 3.2.1 of the UNRES package, specifically UNRES_min with the GAB force field. According to the documentation GAB needs 'export SCCORPAR=$DD/sccor_pdb_shelly.dat' set in the GAB .sh file as a new feature. Not specifying this file gives a file not found for unit 29 error message, and that supports that this file is needed. However when I specify this file, I get an 'Error reading SCCOR parameters' message, and as far as I can see in the following read block in parmread.F, this code cannot possibly read either sccor_pdb_shelly.dat or sccor_am1_pawel.dat. Especially with the shelly file, I really don't know how to modify the reader. Have you ever tested this? Am I missing something?

Thanks,

Istvan

C 5/21/07 (AL) Read coefficients of the backbone-local sidechain-local
C correlation energies.
C
read (isccor,*,end=119,err=119) nterm_sccor
do i=1,20
do j=1,20
read (isccor,'(a)')
do k=1,nterm_sccor
read (isccor,*,end=119,err=119) kk,v1sccor(k,i,j),
& v2sccor(k,i,j)
enddo
enddo
enddo
close (isccor)
if (lprint) then
write (iout,'(/a/)') 'Torsional constants of SCCORR:'
do i=1,20
do j=1,20
write (iout,*) 'ityp',i,' jtyp',j
do k=1,nterm_sccor
write (iout,'(2(1pe15.5))') v1sccor(k,i,j),v2sccor(k,i,j)
enddo
enddo
enddo
endif
C

Hi Istvan,

I guess that you're using the CSA ot MINIM version. Those still have the old torsional potentials so you need the old parameter file rotcorr_AM1.parm for side-chain - backbone correlation. Because you're using the original 1GAB force field, this will not affect your calculations. We'll update the CSA and NINIM sources ASAP, then you'll be able to use the modified GAB force field with new SCCOR potentials.

Best regards,

Adam

Got it. Thanks, Adam.

Best,

Istvan

Adam,

Will you post a note when the new GAB force field is available for minimization?

Thanks,

Istvan

Adam,

I just completed a whole bunch of searches with a helix bundle and found that the current GAB force field gave some clearly wrong, unnatural minima. I can see that PDB based torsion potential could help a lot. Anyway, just please let me know when you had a chance to install the update that you use with MD.

Thanks,

Istvan

Hi Istvan,

Do the artificial minima have very low energies? What system(s) did you try?

I'm not sure if I have already sent you the energy-term weights produced by maximum-likelyhood optimization. They seem to do a good job on helical proteins:

WSC=0.86816 WSCP=0.80317 WELEC=0.37812 WBOND=1.00000 WANG=0.80204 &
WSCLOC=0.47586 WTOR=0.71181 WTORD=0.09999 WCORRH=0.18276 WCORR5=0.00000 &
WCORR6=0.00000 WEL_LOC=0.02888 WTURN3=0.33038 WTURN4=0.40934 WTURN6=0.00000 &
WSCCOR=0.00000 WVDWPP=0.24036 WHPB=0.00000 WSCP14=0.00000 &
CUTOFF=7.00000 WCORR4=0.00000

Adam

Hi Adam,

Sorry for the slow reply, I haven't had a chance to look at this recently. The unnatural minima come with non-native sequences mutating the 1LQ7 structure, so it could be the sequence, that's exactly what I am investigating. I just see such knotted structures but it can be due to the sequence and or the force field. This is why I want to try the "best" force field, so I can blame the problem to the sequence. :)

If I understand you correctly, you want me to try the above weights with the original GAB force field. Correct?

Do you have a tie frame when I could use the new torsion terms as well?

Thanks and best regards,

Istvan

Hi Istvan,

No, these weights pertain to the 1E0LL2Y and not to the 1GAB force field.

Best,

Adam