Reply To: AbinitioRelax under different directories without -constant_seed -jran flag

[Anonymous]

Dear Smileys,

Thank you for your anwsers.

1) What you have done here is not guarunteed to produce identical or non-identical models – it depends on what seed your operating system hands to Rosetta. On many systems, jobs get the system time as their RNG seed, thus jobs started simultaneously will get the same seed. constant_seed + jran can guarantee different seeds (which is what you want). If the models are different, feel free to pool them. You can also check for yourself what the RNG seed was for a run, it should be printed at the top of the log file.

I run Rosetta on a Fedora 13 box. I did not start the jobs at the same time. Therefore, the two predictions might have different seed number. Unfortunately, I did not ask Rosetta to generate the log file, I can not check the real seeds used.

2) All trajectories are independent. So, if it runs through models 1-10, then crashes, and you restart, it checks to see 1-10 are finished and starts at 11. The only relationship between models is that the production of one model moves the RNG along, such that models 1 and 2 and 3, etc, all get different sequences of random numbers.

That means for each model, Rosetta knows when to stop its calculation. Each model came from a completely new try.

3) The error “NANs occurred in hbonding” is serious, but may be related to the other error “ERROR: trying to use a ‘protein’ type silent file for a non-idea l pose core.io.silent: consider using the ‘-out:silent:silent_struct_type binary’ flag”. NANs in hbonding means that something bad happened to atom placement and the hydrogen bond terms tried to divide by zero (or something similar), I can’t diagnose it more specifically. The second error indicates that you’re using a non-ideal pose. Are you starting from a model? Did you idealize it with the idealize application?

I started the prediction with a linear sequence. Here is my option file.
-database /home/zhang/local_programs/rosetta3.1/rosetta_database/
-abinitio:relax true
-in::file::fasta sem1.fasta
-in:file:frag3 aat000_03_05.200_v1_3
-in:file:frag9 aat000_09_05.200_v1_3
-out:nstruct 5000
-out:file:silent sem1.out
-out:path ./

When I added the ‘-out:silent:silent_struct_type binary’ flag, I got the following error:
“ERROR: Option matching -out:silent:silent_struct_type not found in command line top-level context”
Where should I put the flag? What are the essential flags for AbinitioRelax?
How to avoid NANs and the “ERROR:trying to use a ‘protein’ type silent file…”?

4) The “max cycles exceeded” warning is not good, but not terrible. It most likely means that your pose found its way into a region of space where there’s a small error in the derivative of one of the energy functions, and the minimizer couldn’t minimize it properly because the derivative was strange. It won’t cause a big problem. I don’t understand your question about flags.

A long sequence is hard to be folded than a short sequence and Rosetta needs to spend more CPU time to calculate it. My original question was: do we need to adjust -nstrut number only or is there any other flag (or parameter) we should consider according to the different length of the sequences?

Thank you.

XP