[Jansar]

What does "Best RMS" mean (or measure)?

[reader50]

Best RMS, or Best RMSD as the project often calls it.
RMSD = Root Mean Square Deviation

It's a math method used to assign a constant value to a range of varying measurements.

It assigns a value to the imperfection of your modeled proteins. We are working on proteins that have a known shape, trying to perfect the modeling code. The sample proteins are compared to the correct one, and the imperfection value is assigned. Naturally, the smaller the imperfection value, the better.

Once the project comes up with code (or enough crunchers) to reach low RMS values on proteins right away, they will try to crack proteins of unknown shape. That is what they are really aiming at, and attempted to do during the recent CASP prediction contest.

So far, it appears they need to work on the code a bit more. They did not win any of the CASP predictions, and are taking too long to reach low RMS values. That, or they need to add 100x as many crunchers to the Distributed Folding project.

Some further info can be found on their science page.

[Jansar]

Thanks for the link. I was interested by the diagrams on the project's home page where the clear imperfections can be seen.

[Welnic]

Since I had already posted this in a another forum it was a lot quicker to post a second time.

Root Mean Squared is a value used in statistical analysis. You square each number, divide by how many numbers you have so you get the mean, and then you take the square root of that. Squaring the numbers takes care of negative values and means that big deviations carry more weight.

For the series:
0,1,2,1,2,-1,-2,0
square the numbers and divide by the amount of numbers
(0+1+4+1+4+1+4+0)/8 = 15/8
the square root of that is 1.37

Root Mean Squared Deviation is used to compare one series to another.
If this is the original:
0,1,2,1,0,-1,-2,-1,0
and this is test series:
1,0,2,1,1,-2,-1,1,1
then you square the difference between them:
(1+1+0+1+1+1+4+1)/8 = 10/8
the square root of that is 1.12

So when the calculated protein is compared to the actual one, you measure the distance between them at each amino acid, square that and divide by the number of amino acids and then take the square root. Since we are dealing with very small distances you end up with an answer like 8 angstroms RMS. An angstrom is e-10 of a meter.

I don't actually know if the two proteins are compared at every amino acid. They could be compared more than that.