Speeding up FastDesign – Rosetta Commons

This topic has 1 reply, 2 voices, and was last updated 3 years, 7 months ago by Anonymous.

Viewing 1 reply thread

Author

Posts
- April 28, 2021 at 8:08 pm #3760
  Anonymous
```
<ROSETTASCRIPTS>

<SCOREFXNS>

<ScoreFunction name="r15" weights="ref2015.wts"/>

</SCOREFXNS>

<RESIDUE_SELECTORS>

<True name="full_pose"/>

</RESIDUE_SELECTORS>

<TASKOPERATIONS>

<ResfileCommandOperation name="rescmd" command="ALLAA EX 1 EX 2 EX_CUTOFF 1" residue_selector="full_pose"/>

</TASKOPERATIONS>

<MOVERS>

<FastDesign name="design" scorefxn="r15" task_operations="rescmd" repeats="3">

<MoveMap name="mmap" jump="true" chi="true" bb="true"/>

</FastDesign>

</MOVERS>

<PROTOCOLS>

<Add mover="design"/>

</PROTOCOLS>

<OUTPUT scorefxn="r15"/>

</ROSETTASCRIPTS>
```
  I’m trying to design a sequence many proteins of interest. I use the following XML script, but I’m finding that a single sequence design takes 24-36 hours, whereas FastRelax finishes on the order of minutes, as well as the fixbb for sequence design. I’m wondering if I’m doing something wrong, or if there’s any way to speed up the design process.
  
  I’m launching with the following commandline parameters: “-ex1 -ex2 -extrachi_cutoff 1 -linmem_ig 10”
  
  Thanks!
- April 28, 2021 at 9:20 pm #15875
  Anonymous
  Yes, there are many things that can speed up the design process, especially for a large protein:
  
  – Use the Layer selector and the RestrictToResidueProperties or ProhibitResidueProperties task operations to restrict the residue types allowed in core, boundary, and surface positions. You can probably restrict yourself to hydrophobic amino acids in the core and polar amino acids on the surface. This greatly simplifies the combinatorial problem for design.
  
  – Consider turning off ex2 (and maybe ex1), or using an extrachi_cutoff that’s higher. Extra rotamers increase the complexity of the combinatorial problem in a highly nonlinear way.
  
  – Look at your protein and determine whether it can be designed in parts (sequentially), or whether there are regions that you could fix and not design. If it’s a multi-domain protein, for example, you can probably do one domain, then the other, which is much faster than doing both domains simultaneously.
Author

Posts

Viewing 1 reply thread

You must be logged in to reply to this topic.