A compressed representation of multiple-dimensional distribution
by linear base transformation,
and its application to protein 3D structure recognition

Kentaro Onizuka, Tamotsu Noguchi, Makoto Ando and Yutaka Akiyama

Multiple dimensional distribution is represented with fewer number of parameters by linearly expanding the distribution and controlling the cut-off orders of expansion. We adopted this method to the distribution of the relative position between two amino-residues in a protein chain, and applied it to the protein fold recognition problem.

We compared the recognition ratio of three cases, adopting the distribution 1) with respect to the distance (one degree of freedom) , 2) with respect to the 3D position (three degrees of freedom), and 3) with respect to the 3D position and the relative orientation (six degrees of freedom). The result is that the self-recognition ratio of multiple dimensional distribution is far better than that of the conventional distribution with respect only to the relative distance.

Real World Computing Partnership