I am modeling a 422 residue membrane protein sequence using the abinitio protocol provided at this link: https://www.rosettacommons.org/manuals/archive/rosetta3.5_user_guide/db/d38/membrane_abinitio.html. Also, I have attached the files (span, lips4 and flags file) which I am using for generating 10,000 models for the sequence for your reference. Now, when I use the cluster application of Rosetta for clustering the 10000 models, it considers only initial 400 structures. So, I want to know how can I cluster all 10000 models to find the top three clusters? I came across a tool called calibur for cluster large number of models but it has a limitation that I have to extract all the 10000 models in pdb format. Is there any other way of clustering models? I have obtained only a handful of structues with the negative Rosetta score and their energy values are as follows:
SCORE: -16.135 S_00000398
SCORE: -1.912 S_00000663
SCORE: -51.881 S_00000866
SCORE: -73.330 S_00002344
SCORE: -8.932 S_00002716
SCORE: -23.125 S_00003023
SCORE: -1.525 S_00005854
SCORE: -7.247 S_00006069
SCORE: -5.883 S_00006242
SCORE: -67.065 S_00006663
Are these values okay for a 422 aa membrane protein or the flags which I using are not sufficient for generating goor models or whether sampling of 10,000 structures is not enough?
I have also submitted the same sequence to the Robetta server and I was wondering how Robetta models (i.e. flag file options) and clusters the models? Does it generate only 10000 models or more ? I am not able to get the log file for the modeling job submitted to Robetta server, so that I can see how Robetta models and selects the structure; is there any way to obtain that ?
I also have a doubt regarding scoring. While scoring the modeled structures separately using score_jd2 application and score:weights as score3, the total score values in the silent file and the ones obtained through score application have a huge difference. So, what is the difference between the score value in silent file and total_score value obtained thorugh score_jd2 application? I am using the following command for the score_jd2 app:
>score_jd2.linuxgccrelease -in:file:silent Seq1_silent.out -in:file:silent_struct_type binary -in:file:tags S_00000663 -score:weights score3 -out:pdb
So, which scoring value should I use in selecting the lowest energy model?
Thanks in advance for help