So far all my 5.07 CASP6 units are completing. However, one issue bugs me and I'm sure it's not high on the priority list, but here goes. The graphics are not good. Most of the protein on both t209 and t241 are outside the confines of their respective boxes on the "searching", "Accepted", and "low energy" boxes. Also, the analog graphs for "Accepted energy" and "RMSD" are seldom actually in their boxes. This means the "point plot" cross hairs are almost NEVER in the box. Sometimes I see the horizontal line, sometimes the vertical, but almost never both. [edit]for example, one is 68% done, model 5, and the other is 87% model 9, and the right hand corner boxes each have ONE red dot, no teal dots at all.

I'm using 17" displays, one with S3 onboard graphics and one with Asus PCI-express 6200 graphics card.

If someone walked by and asked about the displays, It wouldn't imply a professional image to prospective crunchers.

tony

---

I've noticed that with some of these, the "native" form is all fragmented. Here is an example, I believe it shows Tony's point as well.

---

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/

---

Well, let me tell what I "know", and see if it explains at least the missing red dots. (Note my graphic, how I'm on model 15, but I don't see ANY red dots from prior models).

The red dots indicate the best energy levels of your prior model computations. So, on model 1, there aren't any. On model 2, you should see a red dot for your model 1 prediction... etc. If you get up to 50 models, you'll see many red dots.

The "problem" is that in your current model, your energy readings are scaled dynamically to include this models history. And so frequently, especially early in the model, you don't see all (or any!) of your red dots from your prior models. This is because your current model hasn't come anywhere close to those best energe levels of the prior run, and so essentially the red dots are outside of the current scale of your graph. If you current model improves as the others did, you'll hit lower energy levels and see more of your prior model red dots.

So, it's possible, the current model is just one of many thousands of others that is not producing a great result. Or at least not YET.

---

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/

Well, if properly oriented, some could view this issue as a compliment on how well they've done. I mean, before we were posting about WU that won't finish.
Now that they're finishing, we're posting about something less serious. I mean we're always going to complain about one thing or another. LOL

I guess it's progress.

Well spoken

Anders n

---

Well, let me tell what I "know", and see if it explains at least the missing red dots. (Note my graphic, how I'm on model 15, but I don't see ANY red dots from prior models).

The red dots indicate the best energy levels of your prior model computations. So, on model 1, there aren't any. On model 2, you should see a red dot for your model 1 prediction... etc. If you get up to 50 models, you'll see many red dots.

The "problem" is that in your current model, your energy readings are scaled dynamically to include this models history. And so frequently, especially early in the model, you don't see all (or any!) of your red dots from your prior models. This is because your current model hasn't come anywhere close to those best energe levels of the prior run, and so essentially the red dots are outside of the current scale of your graph. If you current model improves as the others did, you'll hit lower energy levels and see more of your prior model red dots.

So, it's possible, the current model is just one of many thousands of others that is not producing a great result. Or at least not YET.

Well normally the steps start a 0 or somthing like that for each model.

This one does not, it only continues to count right now model 15 step 5250000.

Anders n

---

This one does not, it only continues to count right now model 15 step 5250000.

Anders n

Which result is that?? I can keep my eyes out for it.

This one does not, it only continues to count right now model 15 step 5250000.

Anders n

Which result is that?? I can keep my eyes out for it.

se post 3 in this thread.

Anders n

---

Well normally the steps start a 0 or somthing like that for each model.

This one does not, it only continues to count right now model 15 step 5250000.

Whoa! THAT's not good. Please report the version number the WU is running under.

---

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/

Well normally the steps start a 0 or somthing like that for each model.

This one does not, it only continues to count right now model 15 step 5250000.

Whoa! THAT's not good. Please report the version number the WU is running under.

It is this wu JUMP_ALLBARCODE03_1tul__468_425_0 on this link

https://boinc.bakerlab.org/rosetta/result.php?resultid=18807799

Anders n

---

Well my wu finished and validated ok.

Anders n

---

I wonder.....

The last time I watched the grafics it was on model 15 and

~1,5 H into the run.

How can it report 99 decoys after 2 H then?

Anders n

---

Hiya: Thanks for starting this thread. Let me see if I can answer some of these posts, which are really useful, by the way.

(1) The protein is going outside the box ... we can certainly change the scale on the box! Sorry about that. We can try to make a bigger box for longer proteins. In the future, it wouldn't be too hard to allow the user to zoom in/out on each conformation (note that you can currently do this on the native).

(2) The native protein looks completely messed up sometimes. I've seen this too, and haven't yet tracked down the problem. This might go to lower priority since in a week we'll be running mostly CASP proteins where we don't even know the structure!

(3) Steps are going SUPER high. Oops! That's in the new "jumping" science mode; we'll put in some code to restart the steps afer each model.

(4) Can't see red points from previous models. I think Feet1st is right, but I'll check on it. In "jumping" the red point may never get plotted...

Please continue to post comments on this thread on the graphics. There are some other things we'd like to improve on the graphics, like a description of the protein and workunit that is being crunched, and a more appealing representation of the protein. We're currently suffering a slight shortage of manpower for fixing these aesthetic issues ... that may change in the middle of the summer.

Mine is at 51.5%, 01:45:29, Model 6, Step 2,020,855

---

Well normally the steps start a 0 or somthing like that for each model.

This one does not, it only continues to count right now model 15 step 5250000.

Anders n

They are trying a new technique for model creation called "jumping" It is possible that the counting you are seeing is related to that approach. Part of that routine is the use of information from previous models to alter the path to finding the low energy. It is possible that this has the effect of not resetting the step counter between models, as it is really just continuing to step.

I have seen and reported all the same graphic issues on my systems to the project team. Rhiju has also seen this on his systems and is looking into it.

EDIT: I see Rhiju has already caught this one. I guess it pays to read the entire thread before responding.

---

Moderator9
ROSETTA@home FAQ
Moderator Contact

(2) The native protein looks completely messed up sometimes. I've seen this too, and haven't yet tracked down the problem. This might go to lower priority since in a week we'll be running mostly CASP proteins where we don't even know the structure!

If the native structure is unknown, perhaps you should put a "?" or a smiley or something in the native box. Anything will do.

---