# RosettaCM or FastRelax?

4 posts / 0 new
RosettaCM or FastRelax?
#1

Hi,

I have a Cryo-EM density  and AA sequence for a protein , but the density  is not good enough for assign the residues unambiguously.

I'm particularly interested in one alpha-helix and based on secondary structure prediction, I have roughly known the AA sequence for the helix.

But the secondary structure prediction can't tell the boundary of the helix unambiguously.

Could anybody have any ideas about how I can get the correct register for the helix using Rosetta???

I'd like to try to thread all possible registers onto manually-build poly-A  alpha-helix model and run rosettaCM to refine  each threaded model. I'm wondering  if RosettaCM is a resonable method to do this as the helix is only 20-AA long and for each model it's fully threaded.

Can I just try fastRelax for each threaded model and select the best one?

Thanks a lot,

Tianyang

Post Situation:
Wed, 2020-09-23 07:46
Tianyang89

Hello,

I have actually done this for a few low resolution cryoem structures. I would suggest rosettaCM at the end over FastRelax because when you do these sort of "unnatural" threading minimizations it can help to have rosettaCM to bump you out of weird minima. your method you have described seems reasonable.

To do this I would grab a polyA helix, use that as a template, and use various threadings based on the grishin format (described in the tutorial) to create diffrent "partial threads".  from there you should run rosettaCM with each one separately (~100x) and compare the scores of the top 3.  If there is additional context (ie the helix is interfacing with something )I would suggest adding the context to the partial threads too!

Wed, 2020-09-23 07:54
danpf

Hi!

But I have some questions about RosettaCM.Would you mind giving me some ideas?

1) What does "partial thread" mean? In my case, the poly A template model is 31-AA long and my proposed register sequence is also 31-AA long.There isn't any unaligned region.Is it still called "partial thread" ?

Or should I add more AAs in my regeister?

Here is one example of my alignment file :

#

scores_from_program: 0

0 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

0 KGMRMLEGANINRSLLALGNVINALVSRSKG

2)I understand from literature that RosettaCM stage 1 includes 2 steps:  torsion space fragment insertion and Cartesian space template segment recombination.

My understanding is fragment insertion is only for unaligned region, but not sure if it's right?  Does fragemnt insertion mean de novo building only for unaligned region.

In my case, there isn't any unaligned region, does that mean rosettaCM will not do any torsion space fragment insertion?

3)For Cartesian space template segment recombination, if I only have one template, how this recombination happened?

4) Last question is how rosettaCM can help  to  bump out of weird minima in general?

Thanks again and best wishes,

Tianyang

Thu, 2020-09-24 03:58
Tianyang89

Hello,

How are you setting up rosettaCM?

are you using the python script?  if you are I believe it runs the rosetta application "partial_thread" for you. it makes pdb files that have your target sequence on the template pose

We generally describe unrefined pdb files that have just had their sequence mutated to fit a sequence alignment as "threaded models", sorry if that is jargon!

> Or should I add more AAs in my regeister?

your alignment looks good! you can make many alignments by shifting the bottom sequence to change the register and adding that as a new alignment! (see below)

( i think this is approximately right)

> My understanding is fragment insertion is only for unaligned region, but not sure if it's right?

you are correct but in stage 2 fragment insertion is used to optimize bad geometries so rosettaCM can make more drastic changes than FastRelax

> For Cartesian space template segment recombination, if I only have one template, how this recombination happened?

if there is no other templates this cannot occur

> 4) Last question is how rosettaCM can help  to  bump out of weird minima in general?

Since the fragment insertion will cause larget perturbations to the the backbone, the minimizer will have more starting points that are further from your initial model. This will allow Rosetta to better minimize into the proper minima in case it gets stuck.

## R3 003
#
scores_from_program: 0
0 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
0 KGMRMLEGANINRSLLALGNVINALVSRSKG
---
## R3 002
#
scores_from_program: 0
0 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
0 VKGMRMLEGANINRSLLALGNVINALVSRSK
---
## R3 003
#
scores_from_program: 0
0 AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
0 VVKGMRMLEGANINRSLLALGNVINALVSRS
---


Thu, 2020-09-24 13:44
danpf