You are here

Why are 'fa_rep' and 'fa_intra_rep' higher in 'ref2015_soft.wts' than in 'ref2015.wts'?

3 posts / 0 new
Last post
Why are 'fa_rep' and 'fa_intra_rep' higher in 'ref2015_soft.wts' than in 'ref2015.wts'?
#1

Why are 'fa_rep' and 'fa_intra_rep' higher in 'ref2015_soft.wts' than in 'ref2015.wts'?

So far I thought the soft score function is supposed to feature lower repulsive terms.

ref2015_soft.wts:

# beta_nov15
#   beta energy function following parameter refitting (Frank DiMaio and Hahnbeom Park), November 2015
#
#   This is "soft" version of beta_nov15, equivalent to "soft_rep" to score12, which will be useful when designing at less accurate conditions.
#
ETABLE FA_STANDARD_SOFT
METHOD_WEIGHTS ref 1.34294 3.56185 -2.00840 -2.18603 0.98755 0.74261 -0.09128 2.38347 -0.38872 1.82107 1.87044 -1.32607 -0.30000 -1.50242 -0.28714 0.39411 0.76066 2.59486 1.78282 0.66325 
fa_atr 1.0
fa_rep 1.0
fa_sol 1.0
fa_intra_sol_xover4 1.0
lk_ball_wtd 1.0
fa_intra_rep 0.1
fa_elec 1.0
pro_close 1.25
hbond_sr_bb 1.0
hbond_lr_bb 1.0
hbond_bb_sc 1.0
hbond_sc 1.0
dslf_fa13 1.25
rama_prepro 0.45
omega 0.4
p_aa_pp 0.6
fa_dun 0.7
yhh_planarity 0.625
ref 1
INCLUDE_INTRA_RES_PROTEIN
NO_HB_ENV_DEP

 

ref2015.wts:

# beta_nov15
#   beta energy function following parameter refitting (Frank DiMaio and Hahnbeom Park), November 2015
#
#   Two sets of reference weight are provided.
#      The first is for use in "minimization context" (e.g., RTmin, min_pack, or sidechain relax).
#      The second, and default set, is for use in "packing context" (e.g. Rotamer trials or packing)
#
#METHOD_WEIGHTS ref 1.82468 3.75479 -2.14574 -2.72453 1.21829 0.79816 -0.30065 2.30374 -0.71458 1.66147 2.15735 -1.34026 -1.94321 -1.45095 -0.59474 -0.28969 1.15175 2.64269 2.26099 0.58223
METHOD_WEIGHTS ref 1.32468 3.25479 -2.14574 -2.72453 1.21829 0.79816 -0.30065 2.30374 -0.71458 1.66147 1.65735 -1.34026 -1.64321 -1.45095 -0.09474 -0.28969 1.15175 2.64269 2.26099 0.58223
fa_atr 1
fa_rep 0.55
fa_sol 1.0
fa_intra_sol_xover4 1.0
lk_ball_wtd 1.0
fa_intra_rep 0.005
fa_elec 1.0
pro_close 1.25
hbond_sr_bb 1.0
hbond_lr_bb 1.0
hbond_bb_sc 1.0
hbond_sc 1.0
dslf_fa13 1.25
rama_prepro 0.45
omega 0.4
p_aa_pp 0.6
fa_dun 0.7
yhh_planarity 0.625
ref 1
INCLUDE_INTRA_RES_PROTEIN
NO_HB_ENV_DEP

 

Category: 
Post Situation: 
Tue, 2023-01-24 04:40
JW_24

Soft rep is a softened repulsive, not a reduced repulsive.

The ETABLE FA_STANDARD_SOFT directive in the weights file tells Rosetta to implement an alternative functional form of the repulsive, one which greatly reduces the slope of the penalty as one brings two atoms very close. But the details of this functional form mean that you actually have to upweight those values in comparison to the other terms in order to rebalance sampling such that you get native-like distributions for non-clashing structures.

 

Tue, 2023-01-24 07:51
rmoretti

Thank you very much for this explanation! That makes makes perfect sense.

Tue, 2023-01-24 09:06
JW_24