fixbb design using damped LJ

[Anonymous]

I used fixbb in Rosetta 3.4 to perform fixed BB design using damped LJ potential (by specifying parameter -soft_rep_design). But I noticed that the final score reported by fixbb seemed to be calculated using default weights. (I got to know this by using score_jd2 to recalc the score for deboys using default weights and damped weights and compare them to the score reported by fixbb.) Does this mean my design was really done using damped LJ? Or the parameter -soft_rep_design was ignored and the design was done by using default weights?

It also seems the default weights is “score12”, not “standard” (as the manual claims). What is the background story of these two score weights? Which corresponds to the “HardRep” weights?

Also, what is the difference between soft_rep and soft_rep_design? If one wants to do fixbb with damped rep LJ potential, which should one use?

[attesor]

found this in manual (v2.3)


-soft_rep_design
Specifies the use of an alternate weight set that dampens the Lennard-Jones potential. This option is often used during fixed backbone design since the backbone cannot relax to accommodate small steric clashes.
-soft_rep
Same as -soft_rep_design except that weights are explicitly optimized for side-chain packing with a fixed sequence.


does this mean -soft_rep is more for “relax” with backbone constraints?

[Anonymous]

“standard” used to be standard for non-design things, with score12 standard for things that might involve design. The Rosetta community wants there to be “one scorefunction for everything” as much as possible, so I think most uses migrated over to score12 as the de facto standard, although these is another scorefunction actually named standard. score12 has ~always~ been standard from my point of view, but I am in a design lab…

Soft rep is not primarily a weight set – it is primarily a change to the shape of the repulsive part of the LJ potential. I forget the exact changes, but instead of repulsion ramping quickly to huge values, repulsion instead increases linearly to a low cap as atoms get too close. The idea is to allow small clashes that could be repaired by backbone relaxation. I have noticed soft-rep is a lot less popular in Rosetta3 than it was in Rosetta2 (maybe because backbone minimization is easier?). I do not use it.

soft_rep and soft_rep_design are weight sets and can be seen in database/scoring/weights. They are not that similar. I would guess the one with _design is meant for design, and the other for relax, loop modeling, etc. Notice ETABLE FA_STANDARD_SOFT at the top of the file; this is what enables the differently-shaped potential (although the differing weights matter too).

I have a sneaking feeling Rocco is going to come along and correct me on some critical point I got wrong…

[Anonymous]

-soft_rep_design is not checked by any code which runs when you run fixbb. It appears to be used only in specific protocols but not the core packing machinery. This explains why it didn’t work. (PS, the weights should be printed in the output PDB directly; you can compare those to entries in database/scoring/weights directly instead of rescoring).

-score:weights soft_rep_design is checked by getScoreFunction(), which fixbb uses, and therefore ought to work (and ought not print out score12 scores at the end).

I would guess the design weights are more appropriate than the not-design weights for design work, but only for the obvious reason (there’s no written notes I see on how the weight files were written).

[attesor]

Many thanks!

[attesor]

Many thanks, Steven! I have read Dantas et al. JMB 2007 paper on “DampRep” for detailed changes in the LJ potential. Concerning the application, if one want to do a fixbb design using “DampRep”, should one use -soft_rep_design, or -score:weights soft_rep_design, or -score:weights soft_rep? Rosetta 2.x uses -soft_rep_design. If one uses -soft_rep_design, does Rosetta fixbb design using scoring function with damped LJ potential as target function, or the default score12 as target function? As I asked in the original question, since fixbb reports score12 score as final score, I wonder if damped LJ is used in the design.

[Anonymous]

Looks fine to me.

I also am not sure as to the difference between soft_rep and soft_rep_design. When I’ve talked to people about it, some use one and some use another, without any particular reason given for the choice. Some have explicitly mentioned the fact that they don’t think it matters, because they typically do a hard-rep minimization/scoring after a soft rep packing, anyway.

[attesor]

OK, I will try it myself. Thanks!

Any of the people you asked uses parameter “-soft_rep_design” in v3.x? Is it legacy from v2.x?

[Anonymous]

They use “-score:weights soft_rep_design”.

The “-soft_rep_design” flag apparently is used only in a few specific protocols (enzdes and forge remodel), and is not a general option.

[Author]

Posts