Meet the Latest Protein

[RandomCritterz]

Is any information available about the the latest target? A name, a function, general shape and size?

[Rebels Haven]

A download link?

[pfb]

Protein size is 108...and fairly slow at gen 0 (around 2 hours on a MP2400+)...

[Rebels Haven]

Where did you download it from??? Or is it an auto-update version....?

[pfb]

It auto-updated (via the Daemon setup, not a normal download...) - my *nix box hasn't updated but all the Windows boxen have (nearly an hour ago)...

[jonnyw]

pfb, any chance you could revert to an old copy of a windows client (e.g the one you were using until today, if u still have one) and see if it will update now the main df server is down.

Cos i've got the update files in my webspace, but when i start my client it just says "Server is down for maintenance" even though it should be going to my server and not the df server.

[pfb]

Originally posted by jonnyw
pfb, any chance you could revert to an old copy of a windows client (e.g the one you were using until today, if u still have one) and see if it will update now the main df server is down.

Cos i've got the update files in my webspace, but when i start my client it just says "Server is down for maintenance" even though it should be going to my server and not the df server.

sorry - don't have an old version of the Windows client...have just noticed that my *nix client has updated as well so not sure why yours hasn't updated...

[Rebels Haven]

Hmmm... my clients are just buffering....

[Stardragon]

The new protein is called 1YCC and comes from baker's yeast.
It has 108 residues. We have worked on this protein in the past, and now would like to get results for it with the new algorithm.

Since it's almost twice as long as the previous protein we worked on, it will take a while longer to crunch through

[RandomCritterz]

Originally posted by Stardragon
The new protein is called 1YCC and comes from baker's yeast.

Yeeha! Something new to google for, instead of working!

I wonder, if I crack open a beer will I be able to taste the 1YCC?

[[DPC]Mobster]

What's our goal in terms of energy, RMSD and amount of work te be done?
What was our best result using the old algorithm?

[Grumpy]

This new Protein is the Son of Satan..it keeps getting stuck in loops on 3 different machines with different OS/Hardware configs..the only clue is one machine keeps spittting out "[002.002] {buildit.c line 256} Cannot set co-ordinates" I think the Yeast is fermenting

[pfb]

Originally posted by Grumpy
This new Protein is the Son of Satan..it keeps getting stuck in loops on 3 different machines with different OS/Hardware configs..the only clue is one machine keeps spittting out "[002.002] {buildit.c line 256} Cannot set co-ordinates" I think the Yeast is fermenting

Just looked at one of my error.logs and saw 'Tue Oct 28 20:41:31 2003 ERROR: [002.002] {buildit.c, line 256} Cannot set co-ordinates' - but the client has continued so not sure if it is a critical error or not...

[iggy]

Same thing here, but only on one systems, all others are behaving.

Just had two generations that took 2:50 hours to complete (altogether) on another system, and energy is fluctuating from 40 to 100...

[Grumpy]

All those with Overclocked Boxen, watch your Temp please. This Client has 2 of my Boxen 6 degrees hotter than what they normally run at, they seem to go into loops and hammer the cpu and it gets very hot before it reaches the next part of the algorithm where the temp drops again. If it gets to a tight spot at this point


Got the extinguisher ready


Well, got a few Fatal Errors on 2 Boxes..Win XP Pro and Win2K. My Clients have updated again in the last dew hours, wot da hek iz that about ?? So far, I rate this Protein 0.5 out of 10. It gets the 1/2 point because ALL my Systems auto updated for the 1st time since I started this project

[Rebels Haven]

1 out of 18 got caught in the "Heat Loop" here....

[Keller]

There are some informations abput the new Protein:
http://www.rcsb.org/pdb/cgi/explore....57001067368462

[rsbriggs]

That's one twisted protein.....

[Grumpy]

"Danger Will Robinson"

Check your Ram Usage please ...sometimes after a restart the CLIENT DOES NOT USE EXTRA RAM DESPITE WHAT YOU DO. I had to restart 4 times before it used the extra ram..what the hell gives with this Client...Now I am abolutely Grumpy, my 6 Brothers are coming to get you when the shift at the Diamond Mine in finished :bs:

[rstarr]

This is one bad moejoe. Everyone has been humbled by this new work. It seems my farm system went on a holiday. 5 of my machines are just looping like everyone else. Temps are high, work is very slow, results even slower.

At this rate it'll take over a year to make the target rate of 5 billion.

Methinks it's time to put this yeast to rest. RIP

[RaginSteveK]

one or other of my boxes gets hung [ I assume- completing 5 generations overnight !! instead of 57, like in the good old days !]..

but when I look at my best structure, to date-- and it's EARLY in the game, I'm impressed !!!

thinking that such a twisted [ perverted?] beast may cause real problems for simply any old algorithm ...

be willing to bet we see some procedural tweaks down the road...
starting with better scoring, for the stats ho's among us...

my machines are crunching hard, and getting no where, points -wise !!

[RaginSteveK]

just a thought...

but it would probly be good to hear something uplifting from that Hogue guy..

to remind the troops once again that more is happening here than overheated processors and inflated utility bills...

[jonnyw]

Originally posted by Grumpy
[BCheck your Ram Usage please ...sometimes after a restart the CLIENT DOES NOT USE EXTRA RAM DESPITE WHAT YOU DO. I had to restart 4 times before it used the extra ram..what the hell gives with this Client... [/B]

is this not just when its between generations calculating the trajectory thingy ma bob.

i've noticed mine doing this on previous proteins, and when left to run regained full ram usage after a while.

'tis damn slow though

[Grumpy]

Nope, wemt through 4 Gens at 20 MB the whole time, fluctuating between 16 and 23 MB

[Blackeagle]

The largest concern for everyone should be that a large number of users taking part in this are really unhappy with the new scoring arrangements.

It's not that they need to see higher numbers per day to be happy, it's that this much slower scoring system gives a huge advantage to teams or individuals that have compliled a lot of points under the old higher scoring method.

I think it may well have been better recieved had all scoring gone back to zero. The fastest individuals and teams would still rise to the top. And newer users who came along with very fast rigs would have equal chance to reach the top, in a reasonable time. But to reduce scores per day by 2/3s is to inshrine the teams and individuals on the front page there for good. Very discouraging to new folders, why should they bother? There are other distributed computing projects that will not subject them to such a penalty.

I'd suggest wipping out all scores at the start of the next protien. And also let everyone know this is to be done and why, so they can offer comments if they wish.

Loss of users taking part in ANY significant numbers at all is bad. And I'm concerned that is what may happen. And the medical research is to important, I hope we can avoid that.

[HaloJones]

Blackeagle,

This very same argument went on during Phase I. Yes, this protein is slow, but I didn't see anyone complaining when the last one was fast. The fact is that with DF, we get the same points for a slow protein as for a fast one but everyone scores at the same rate whether the protein is fast or slow. It will take me longer to catch the people ahead of me and longer for those behind me to catch me with a slow one and the opposite with a fast one.

While restarting every protein may encourage new starters, it also upsets those with prior stats. When the scoring was reset for Phase 2, a fair number left and never returned. I have no evidence but I believe the downward trend in participants has been so for a long time.

It is unfair to reset and unfair not to. Unless the scoring is going to be weighted according to the size of the protein, whatever happens will be unfair.

I lost nearly 900 million points when the points were reset. Tell me again about how fair it is to restart each time.

[isp]

I agree with BlackEagle. I just started a few weeks ago and was really enjoying the project. Now that the scoring changed with a new protein it has me totally turned off. I think I'm going to go back to f@h were I don't have to worry about them changing the scoring system so much

[willy1]

I don't believe anything in the scoring system has changed. Same points for each generation as the last protein.

This protein is slower. The next one might not be. Every protein folded is different.

The only time the stats were reset was when the scoring system DID change, between Phase 1 and 2.

[pfb]

the 'points scoring' hasn't changed but the 'energy scoring' has...maybe people are getting mixed up with the 2?

[tpdooley]

Blackeagle:
If you've watched the active user's number over the past year, you'd notice that the number plummets after the changeover every time. There's evidently lots of people on vacation, on dialup lines, or who just aren't paying attention and thus don't put any special effort into getting the system switched over to the new protein/new algorithm. It usually takes a few weeks to get back up to the same number of active folders. And much longer when we switch from a speed demon protein to a slow one.
We've also got a number of people that are waiting for the new algorithm to get near the error free status of the client we were using at the end of Phase I. We seem to be getting pretty close.. or those who are having the problems aren't taking the time to report them in a way that will allow them to be fixed - expecting everyone else to do it. Please post the errors you see with as much detail as possible - especially with the steps used to reproduce it, and as much detail about the system having the problem so that Howard can recreate it in house and then cure the problem.

As for the relative unfairness of people having a headstart and a faster client - I started this project with 1ea Athlon 1800+ (stock speed) /w 512 Megs running win2k. The 3rd or 4th place folder at the time had quite a few months head start, plus 100 each Athlon 1800+s (overclocked) /w 256Megs running a flavor of Unix. When Phase I ended - what was the ratio of her score to mine? roughly 1:1. (She gave up after awhile.) Perseverance pays off.. (as does increasing your number of folding machines I may be up to 20 folding machines; although most of those are part time. My six dedicated folding machines are full time folders..


----
For Howard or Elena: What were the first week scores (rmsd) for this protein under phase I and the previous scoring algorithm in Phase II? And what did each end up with at the end of the folding?

[Stardragon]

As for resetting the scores, there has been much discussion about such an approach in the past (around the switch to Phase II) and as per the responses, most users would be increasingly unhappy if each protein would reset the score. Keep in mind that distributed folding seeks to understand how proteins fold, and if this were easily done with tiny proteins, we would be much farther ahead by now. It just so happened that the previous protein was unusually small and fast, but the proteins we work on are known to fluctuate somewhat in size and speed - it's an up and down process, so it is best to leave things as they are.

[Brian the Fist]

We have not changed the scoring method (and zeroed the stats) since the start of Phase two several months ago, when it was announced far in advance. We will not be changing them again unless a significant change to the algorithm is made. Everybody has a different idea of fair, so we went with what works easiest for us. If you don't like the way our scoring system works then no one is twisting your arm to participate but we'd really prefer people help for the science anyways

[FoBoT]

Originally posted by Brian the Fist
If you don't like the way our scoring system works then no one is twisting your arm to participate but we'd really prefer people help for the science anyways

but he lets us moderate stats whores run DF anyway

i know the protein speed changes are a shock to those new to DF that don't expect it, but it all works out over time. people move boxen on/off DF due to many reasons, one is the protein speed. i see it as making the project more interesting (from a DC stats perspective, not from the "science" perspective) , it means you need to use some strategy as to how you run your boxen (for those that do multiple DC projects)

its a Good Thing™

[shortfinal]

I've noticed on the team statistics page that it lists the total # of points for the current protein for each member on the team. Since this is being recorded, how about on the High Flyers page a top 10 list for the current protein?

Shortfianl

[isp]

Personally I think there should be two scores. A total score with every protein you've ever worked on included in the total. A current score just for the current protein. Once a protein no longer need to be worked on and they switch to a new one your ammount of points earned for that gets added to the total score. Otherwise the speed varience between proteins doesn't reflect a true score.

[Paratima]

Originally posted by FoBoT
but he lets us moderate stats whores run DF anyway

Moderate!!

[jonnyw]

i dont think the scoring should be re-set at each stage, but i do think it should be weighted, so 24 hours of folding always gets approx the same score.

I was getting around 150,000 points a day on the last protein, and since the new one started last tuesday I haven't even reached 100,000 yet.

Its certainly put me off leaving my computer on as much, as it hardly seems worth the effort.

AS i said above I think the scoring should be left alone for fast proteins, and weighted up for the slow ones.

That way the stats junkies would be happy with all the points they're getting and the ones in it for the science wouldn't lose out either.

Just my opinion though....

[isp]

I guess I'm a stat junkie :

It's okay though right?

[Blackeagle]

Originally posted by pfb
the 'points scoring' hasn't changed but the 'energy scoring' has...maybe people are getting mixed up with the 2?

Exactly! !

And I'm one of them. If there is NO change in the scoring system what does the above refer to then?

And did this energy scoring change create part of this dramatic slow down in scoring?

If the only change is due to this protien being slow, fine, just have to wait for the next fast one. But I hope this energy scoring change isn't part of the large slow down in the scoring seen.

I do know this, I've seen one pretty strong folder go back to folding at home due to this.

[jonnyw]

i'll try to answer this (but am not necessarily saying that i'm right )

Originally posted by Blackeagle
[B]Exactly! !

And I'm one of them. If there is NO change in the scoring system what does the above refer to then?

I think some people are talkin about the "best energy" scoring of each structure folding - This used to be RMSD (and compared to somerthing they knew worked), and is now not compared to anything cos they don't know how it should turn out (they just wanna see what "best energy"scores they can get).

Others (myself included), have been talking about the points that us (stats hoes) get for each generation we upload, which is always ~130,000 points per 250 generations uploaded.

And did this energy scoring change create part of this dramatic slow down in scoring?

If you're refering to the scoring of the lowest struct ("best Energy"/RMSD type thingy) (which can be found on the "high flyers" page), I have no idea

If by scoring you are refering to stats scoring, they've slowed down cos the latest protein takes longer than usual to fold, and so it takes longer per gen, therefore less points to the user.

hope i've answered your questions (vaguely) and not confused you too much



otherwis i'm sure someone else will be able to explain it all much better