Hi,

What does SAVE_ALL_OUT and IGNORE_THE_REST in the WU names mean?

Christoph

Each model starts at a random point on the protein at a random orientation and crunches from there. How it crunches is based on how the task was built in the Bakerlab. It is sort of like an acorn. If you could program an acorn to tell the tree how to grow, and how to react to various environmental conditions it will encounter through it's growth as an oak tree.

These are references to how the models are run, sort of like how the tree of your acorn will grow. And I believe it is basically a reference to which branches of the growing tree they decide to keep and which to sacrifice when the tree is unable to sustain all of it's branches.

By trying different approaches, they are learning which produce the best models, and which are able to run with less computer power.

---

Add this signature to your EMail:
Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might!
https://boinc.bakerlab.org/rosetta/

Each model starts at a random point on the protein at a random orientation and crunches from there. How it crunches is based on how the task was built in the Bakerlab. It is sort of like an acorn. If you could program an acorn to tell the tree how to grow, and how to react to various environmental conditions it will encounter through it's growth as an oak tree.

These are references to how the models are run, sort of like how the tree of your acorn will grow. And I believe it is basically a reference to which branches of the growing tree they decide to keep and which to sacrifice when the tree is unable to sustain all of it's branches.

By trying different approaches, they are learning which produce the best models, and which are able to run with less computer power.

That's beautifully said.

They're internal codes to tell the server how to treat the results.

Rosetta@home has the ability to produce a huge number of decoys, which can take up a large amount of disk space. Often we're not interested in the exact structure of all of results - just the best scoring ones. So by default, we only save the complete structure for the best 5% or so decoys, and just store the scores and other simple information for the rest. The SAVE_ALL_OUT tells the R@h machinery that it should keep the structures of all the models, not just the top 5%. (Don't worry - you get credit for all the decoys you produce, not just ones in the top 5%. The cutoff is just there to keep from filling up server disk space with models no one will look at.)

The IGNORE_THE_REST is for job management. When the server is combining the results from different work units, it bases it on job name. However, it first strips off the IGNORE_THE_REST and anything following it from the job name. This way you can have a number of separate jobs which are treated as a single job for the purposes of result accumulation. (Again, this is all internal - there should be no visible effect on your end for jobs with/without IGNORE_THE_REST in the name.)

Feet1st is correct in that R@h takes random starting points, and different models stem from different starting points and different paths of growth, but that's really controlled by other mechanisms, rather than SAVE_ALL_OUT and IGNORE_THE_REST.

I've always wondered this. Thanks!

---