Text-based client display

[Mikus]

Question - is the colorful ascii picture drawn by the text-based client a 'random' drawing, or is it representational of the structure the client is constructing ?

The new client will from time to time display the message "caclulating energy" -- but I note that usually the accompanying ascii picture does *not* look "compact" to me.

mikus

[runestar]

From what I understand its an abstract representation of the protein. I don't know if the colors and symbols are random or not. Some of the symbols look intentional like: - \ / |

Those seem to match up when the drawing changes direction.

RS½
PBB

[Brian the Fist]

The ASCII art is indeed random however it corresponds crudely to what a protein may look like, in 2D. The OpenGL screensaver shows the REAL structure it is building in memory..

[runestar]

So basically it corresponds to crudely looking at the protein without depth...

RS½

[Mikus]

Originally posted by runestar½
So basically it corresponds to crudely looking at the protein without depth...
RS½

But then may I ask again -- what is an "efficient" structure supposed to look like ? In particular, is it ok for its ASCII 2D image to stretch halfway across my screen ?
mikus

[MAD-ness]

If they knew what the structure looked like, they wouldn't need us (or this client).

Because we are working with proteins of unknown structure (shape/geometry) here there is not a base-line to compare it to. What they are doing is taking known behavioral trends, probability maps, etc. and predicting what the protein shape will look like in 3d space, based upon the behavior of other known proteins (especially other known SIMILAR proteins). There are base proteins which are incorporated in larger proteins (a certain string of amino acids in a certain shape that is repeated in larger strings of amino acids) and if you can identify these then you fill in a lot of the blanks, but you still have to "predict" the remaining portions of the amino acid chain.

It is late, I am tired, I suck at science (and math) so please forgive me if this is inaccurate or incomplete, but it should convey the general idea.

I suggest the "science of distributed folding" I think is the name of the thread on the Educational forum. All of this stuff is visited in detail.

There is no way for ANYONE to take a prediction for a protein of unknown structure and judge with complete accuracy how close that protein's structure is to the "true" structure. Until the protein's structure is discovered (via x-ray crystalography and other lab work) there is nothing to compare against.

One of the things that the guys running DF are trying to improve are the methods they use for analyzing a structure prediction and judging the accuracy of it (without knowing the true structure).

How can you grade your own test if you don't have the correct answers? Pretty damn tough, but if you look for trends, analyze the behavior of the test taker, etc. you can eventually make some informed guesses. As time passes and finally see the correct answers, you go back and then can analyse not only how you screwed taking the test (i.e. coming up with the answers) but also how well you did in figuring out whether or not you did well on the test (without knowing the correct answers).

Perhaps a bad analogy but maybe it conveys the difficulty (immense) of the challenges involved in what I think they call "ab inito protein structure prediction".