You are here

Abinitio the Broker protocol

2 posts / 0 new
Last post
Abinitio the Broker protocol

Hello everyone,

I am reading about the Abinitio protocol to see how I can include a checkpoint file in the Abinitio flags, but i could not find any information on how to input a checkpoint file. so this is my question:

  • can I input a checkpoint file into Abinitio? if i can what is the flag     -in:file:  ?


While reading I found something called Abinitio using the Broker. I tried to read more about how different is normal Abinitio with regards to Broker Abinitio, but I could not find the information, only the tutorial. So my second question:

  • what is Abinitio using the Broker? is it supposed to be a better Abinitio? or for specific protein families (like membrain proteins)? is it a new standard?


The broker tutorial says i must have a checkpoint file and a jufo_ss file, but there are no flags for them in the options files, how are they inputed?

Post Situation: 
Sun, 2019-03-24 03:56

Regarding checkpoint files, there's two types of checkpoint files that are relevant to abinitio.

The first is checkpoint files which the ab initio protocol itself can create (depending on details of how you're running it) as a "resume from here" marker.  You shouldn't ever need to explicitly specify this as an input -- if the ab initio uses an intermediate checkpointing file, it should use it to resume the computation when you re-launch the job with the original command line for the job which made such a checkpointing file. (This is something of a general thing -- most of the time, if possible, Rosetta should be able to pick up from where it left off if it didn't completely finish the run.)

The other checkpoint file is an output file from BLAST runs which Rosetta can use for various profile predictions. This file is used primarily in the fragment picking process by Rosetta. It isn't really used directly as an input for ab initio runs.

That's most of the time. This is where the difference between the various ab inito protocols come in. The broker is something of an extension to ab inito, which allows for more involved protocols. Instead of a simple ab initio protocols, you can add additional features ("claimers") which can influence how the ab inito protocol works. But these claimers (and the input files thereto) aren't specified directly on the command line. Instead, they're typically specified in a "topology broker setup file" which is a text file that's given on the command line to the broker. 

I don't have much experience here, but that's my understanding of where the BLAST checkpoint file (and jufo_ss file)  used in non-fragment picking sense comes in. There's some claimers which can use the blast checkpoint and jufo secondary structure prediction files to add constraints/biases to the ab inito protocol, alterning the ab inito trajectory to "better match" the data in those files.

Whether that's worth it or not is another question, and is likely system dependent. Often the information from the blast checkpoint and secondary structure predictions are adequately captured by the fragment files, and attempting to also include them during the fragment assembly stage is another layer of complexity which doesn't substantially improve the structure prediction. But if you have a particularlly recalcitrant protein, there is a possibility that using the more involved broker protocol might help.

Fri, 2019-03-29 09:20