PDA

View Full Version : store just "best" protein?



wirthi
06-11-2002, 03:13 PM
Hi,

does the server store just the "best" structure of each user? Is this - besides stats - the only thing the server saves?

If that's the case, WHY only the best structure? I mean, it't nice that you store MY best protein, but for example Howard computes proteins 200 times faster than I do - It is quite likely that Howard's second best structure is better than my best I guess (perhaps even his 200th-best - don't know, I am no maths-mastermind).

So, if you send in the five structures to CASP (as I have read somewhere), then all five have to be from a different user. What, if Howards second best strucutre is in fact better than the fifth-best structure we send in ....

I assume the probability is very low that this happens, but I am a bit concerend about wasting good folds :bang:

Greets,
Wirthi
Team Austria

Stardragon
06-11-2002, 03:41 PM
You are correct is saying that from all the structures generated by one user, only the "best" structure gets submitted from every set, and the overall "best" structure the user ever generated is stored on the server.

Although it is theoretically possible that my 2nd best structure is better than your best one, there are a number of things to keep in mind:

Even though I may be generating many more strucutres, the percentage of the conformational space I sample would still be reasonably small, so more structures generated does not necessarily translate into a greater number of better structures, at least not on the scale of comparison in question.

Ultimately, although we can submit 5 structures per CASP target, we are interested in the overall best structure we can find.

Although the possibility exists that second or third best structures may be of some use to us, server space is limited so we need to stick with the current setup.

I hope this clears things up :p

wirthi
06-11-2002, 04:21 PM
Hi,

yes, it does, but you could for example keep a list of the 5 (10, 100, 1000, whatever) best structures besides the list of each user's best structure.

You write it is very unlikely that a user can place two structures in this top-5 list. I also think it is almos impossible, but at least is IS possible. If it's worth while implementing, i can't estimate (I don't know how you use this five best structures; can I read from your statement that you only use the overall best structure - then all the discussion is useless anyway?!?!)

Greets,
Wirthi
Team Austria