Folding Forum

While rummaging around github (where the source code for FAH is kept) I found a new version of the viewer.

I spun it up and it shows the protein starting of as a big lump then splitting in half - both halves continuing to fold.

Can this really happen or more likely a viewer issue? If it's the latter then I'm disappointed it isn't the former - far more interesting!

CPU Project: 13850 (Run 0, Clone 10448, Gen 16)

https://www.techpowerup.org/edit?file=/ ... 4d04cf0949

Here's an explanation I got from the project owner :

justinrporter wrote:It’s almost certainly the periodic boundaries. We simulate not a single protein in a box but actually an infinite lattice if identical proteins only one of which is represented explicitly. When the protein reaches the end of the box, it just enters back in in from the other side.

iirc Asteroids had a very similar feature

Neil-B wrote:iirc Asteroids had a very similar feature

Thanks for reaching out toTOW and explaining. This combo of science/software is just brilliant.

The split occurred somewhere between snapshot 1 and 2.
Stayed separated for ages then like gravity joined up again in the center, coming toward each other from the edges.
Only to change their minds and fly off again.

Best film I've seen for ages!

NuovaApe wrote:Thanks for reaching out toTOW and explaining. This combo of science/software is just brilliant.

The split occurred somewhere between snapshot 1 and 2.
Stayed separated for ages then like gravity joined up again in the center, coming toward each other from the edges.
Only to change their minds and fly off again.

Best film I've seen for ages!

If you're talking about the game, then maybe that's true. If you're talking about the atoms in FAH, then there is no gravity because the segments are actually joined except that an edge of the box transforms them in the image by the exact distance between the two edges of the box.

To get a viewable image, some work is done by the FAHCore and some is done in the viewer. Neither codebase is exclusively responsible for the entire process. Which half of the process SHOULD be responsible for which step should be negotiated by managment and the assigned teams need to follow the rules.

As long as the test case never moves across the box boundary, the team can claim that their code is correct because nothing is split, but that's obviously not the right answer. Somebody needs to transform the disconnected pieces into a single, connected protein image.

Hi Bruce. I was talking about the 2 "cleaved" halves coming back under gravity/strong/weak magnetic forces. That's how it appeared at least. I'm no physicist so they all look the same to me;-)

They didn't go out one side of the box and come back in the other like asteroids or some universe physics theories.

They rotated "to fit" when attracted then repulsed. But that's just the snapshot sequence playing backwards/cycling methinks.

TBH I'm not bothered about viewing. It's just I've never seen it work on Windows and some guy shared a build that let me peek.

Seeing it for a brief while let me understand, visually, the complexity of this stuff. All those parallel dynamics that nature does with ease.

It left me understanding less.

The purpose of all this is not just to see start and end, but all the steps in between, right?

How does one analyse all these "frames" to identify opportune moments for mending/breaking a fold sequence.

Watching the viewer "video" you see patterns, circles of the same color, places where you could poke a stick.

I don't get how the results can be analysed computationally. SETI has easy to detect spikes. What are yours?!

Please don't answer. I'll end up understanding even less until I'm in negative knowledge land;-)

NuovaApe wrote:...The purpose of all this is not just to see start and end, but all the steps in between, right?

This is what F@H sees via simulation:

NuovaApe wrote:...How does one analyse all these "frames" to identify opportune moments for mending/breaking a fold sequence...

That would be the scientists and researchers using statistical sampling to find the most promising aspects of the protein folding to analyze.

NuovaApe wrote:...I don't get how the results can be analysed computationally. SETI has easy to detect spikes. What are yours?!...

It's molecular dynamic (https://en.wikipedia.org/wiki/Molecular_dynamics) within the protein folding system (https://en.wikipedia.org/wiki/Protein_f ... in_folding).

Thanks for that PantherX - my brain needs a detraction from the obvious right now.

That "squiggle" diagram looks like an old Richard Feynman squiggle diagram I never mastered either.

I stand on the shoes of giants, always looking up;-)

Can somebody send back the an example a copy of viewerFrame%d.json and viewerTop.json files when this happens?

Bruce - lightening never strikes twice and I've viewed/overwritten json since.

It appears lightening does strike twice and I'm seeing same again (CPU).

I have an unofficial build of Windows viewer and don't want to waste peoples time on this. (meertronix)

I've paused.

There's json and the argonoughts in my work folder.

Will zip and unpause...

Please don't spend too much time on "viewer" problems for me. I wanna crack on and munch numbers;-)

NuovaApe wrote:...I have an unofficial build of Windows viewer and don't want to waste peoples time on this. (meertronix)...

My take is that it's a data gathering exercise. Regardless of if it's a bug or not, we have the data to analyze either now or in future. Given that it's rare to reproduce, better to get it sooner than later when reliably reproducing this issue can be a problem.

The split protein is a long-standing issue. It only strikes when there's enough charge in the clouds to create the bolt, but I still think it should be fixed. You're not the first one to run into it and you won't be the last.

It's not that rare - seen it occur with 3 recent WUs (CPU). The last one had 3 splits in the same WU. I took an mp4 of it.

I've backed up the viewerTop/viewerFrame json files for this last one. How best to "send back" these files?

Presumably you don't need all viewerFrames, just the snapshot number when the split occurred?

bruce wrote:Can somebody send back the an example a copy of viewerFrame%d.json and viewerTop.json files when this happens?

Happy to Bruce.

PantherX wrote:

NuovaApe wrote:...I have an unofficial build of Windows viewer and don't want to waste peoples time on this. (meertronix)...

My take is that it's a data gathering exercise. Regardless of if it's a bug or not, we have the data to analyze either now or in future. Given that it's rare to reproduce, better to get it sooner than later when reliably reproducing this issue can be a problem.

bruce wrote:The split protein is a long-standing issue. It only strikes when there's enough charge in the clouds to create the bolt, but I still think it should be fixed. You're not the first one to run into it and you won't be the last.

I've seen the split in 3 WUs in as many days. The last had 3 splits in the same WU. I've backed up the viewerTop/viewerFrames for this one, and took an mp4 of the viewer splitting.

Let me know how to send these files back; FTP, paste here, or pastebin.com - not email as having troubles with it.

Do you need just the viewerTop + viewerFrame%d for the integer snapshot number when the split(s) occurred?