A compressed representation of multiple-dimensional distribution by linear base transformation,
and it's application to residue-pair-relative-distribution of proteins

Kentaro Onizuka, Tamotsu Noguchi, Makoto Ando and Yutaka Akiyama

We propose a representation method to strongly reduce the number of parameters representing a multiple-dimensional distribution. The method reduces the number of parameters by linearly expanding the distribution with orthonormal linear bases.

To evaluate the representation performance of the method, we apply it to the distribution of relative position of two amino-acid-residues in a protein chain. We firstly apply this method to represent the distribution of three dihedral angles (φddd) which almost perfectly represents the relative position of adjacent residues, and then apply it to the distribution of the relative position of two residues separated in the sequence, where the relative position is represented by polar coordinates (rppp) and Euler's angle (θeee).


Real World Computing Partnership