I have a question about clustering. I'll present my problem first.
I have a simulation of a protein trayectory (a file with a time series of protein 3D-coordinates), from it I can calculate the Root Mean Square desviation (RMSD) beetween any pair of structures in the trayectory, which is a rough idea of 3D similarity beetween them. If for every conformation (trayectory frame) I calculate the RMSD with all other frames in the trayectory, I get a RMSD matrix, which is the data i have now.
In some published works a conformational analysis is performed with
this. They represent every frame of the trayectory as a point in the
space (or plane); and having N(N-1)/2 independent distances (RMSD
matrix, D[i, j]) they arrange them in space in a way that the real
distance d[i, j] in the plot would be very close to the Matrix
distance, so the residual 'e'
N e = SUM( | D[i, j] - d[i, j] | ) i, j; i<j
I have attached a .jpg file with an example taken from a paper. It would be also usefull for me to link points beetwen them acording with time, to have an idea of the temporal changue in conformation.
I read the class and cluster package, and i did not find any function able to do this, and I cannot imagine how to use existent funtions to obtain results i want. I would like to ask, according to what i expose above, whether is posible to obtain a graph like that in R. I Know from manuals that in R is not posible to interact with graphs, So if it would be posible to obtain that graph, there would be difficult to zoom on it, or to know the identity of a particular point by clicking. Nevertheless i read that there is other programs that can be used to obtain movies of R data, or interact with graphs ¿Someone knows their(s) name(s)?
Any help or comment would be very useful
This archive was generated by hypermail 2.1.8 : Sun 23 Oct 2005 - 15:06:30 EST