You are here

abinitiorelax: No funnel

7 posts / 0 new
Last post
abinitiorelax: No funnel

I'm attempting to do an ab initio prediction. I have created 360.000 structures using: AbinitioRelax.linuxgccrelease with the following flags:

-in:file:fasta inputs/mINT.fasta
-in:file:frag9 inputs/aat000_03_05.200_v1_3
-in:file:frag3 inputs/aat000_09_05.200_v1_3
-out:file:silent OUT
-out:nstruct 1
-database /x/Rosetta3.1_Bundles/rosetta_database

I assume that the rosetta energy score is the one in the first column of the score.fsc file just after "SCORE: ". All the scores I get are positive?!

I have identified the lowest scoring file (score = 1.7) and calculated RMSDs of all other structures compared to this. When making a plot of the (assumed) score as function of RMSD I get cloud but no funnel!?

I hope you can help!

Tue, 2010-07-27 06:14

Caveat: I don't use ab initio, so I'm guessing.

A) I'm not sure what happens when you use both out::file::silent and out::pdb at the same time. Are you getting both PDBs and a silent file? and a score file too?

B) Are the columns not labeled in the score.fsc file...? What do the first few lines look like? I'm not familiar with .fsc, is that .fasc?

C) It's perfectly possible you have a cloud with no funnel, it happens. What's the spread in values? If the range is 1.2-2.3, then it's probably not rosetta total score in that column. If it's 1.7-500, then it probably is.

D) It's possible to get all positive scores. It means Rosetta hates all the models it made.

Tue, 2010-07-27 08:40

First line in scores.fsc is the sequence. Second line:
SCORE: score vdw cenpack pair env cbeta rg hs_pair ss_pair rsigma sheet clashes_total clashes_bb time description

This there any documentation on the output to this file?

My scores range from 1.7-194.

It seems unlikely that running more predictions will fix the problem! Is the any way to get Rosetta to search other conformations or must I conclude that rosetta cannot to the prediction of this 160 aa protein?!

Wed, 2010-07-28 02:02

The first issue is that you have centroid poses, not fullatom poses. It is expected that centroid poses have positive scores. I thought you had fullatom poses. So, at least the only-positive scores aren't a worry.

You might be able to try running abrelax with a pre-supplied secondary structure prediction, that might improve Rosetta's results. You could also try running the post-abinitio relax phase (that's what it's abrelax) and/or getting fullatom results, but I don't know if that will help any.

Wed, 2010-07-28 07:39

I realized the missing relaxation and did fix this. This resulted in all negatives (range -330 to -180). But still no funnel :(

Is there any chance that changing from centroid poses to fullatom poses will fix my problem?! In any case: How is this done?

Thank you!

Sat, 2010-07-31 03:37

The manual doesn't say. You can try the -help option; I believe the abrelax authors liked that particular documentation system (as opposed to the manual) and it may offer more advice.

Sun, 2010-08-01 12:23

>You might be able to try running abrelax with a pre-supplied secondary structure prediction, that might improve Rosetta's results.


how do you give a secondary structure prediction to AbinitioRelax.linuxgccrelease? it is not obvious from online manual nor "-help" command option.


Thu, 2010-08-05 13:28