Improvement of protein secondary structure prediction
using the 3D-1D compatibility algorithm

Tamotsu Noguchi(1), Masahiro Ito(2), Yutaka Akiyama(1), Ken Nishikawa(2)

Recently, the methods for 3D-1D compatibility algorithm (threading) has been extensively studied, as the Protein Data Bank (PDB) entries has been increasing rapidly by the improvement of X-ray crystallography and NMR experimental techniques. On the other hand, a common feature of the protein secondary structure prediction algorithms was a limited strategy taking only the local information along the chains (typically within a window 7-19 residues long) into account.
Ito et al. have proposed a method for the prediction of protein secondary structure using the 3D-1D compatibility algorithm (SSThread), which relies totally on the global aspect of protein. The average prediction accuracy for proteins in their test set was about 69% in the three-state assessment of α, β and coil.
We have developed the SSThread and reconstructed the 3D profile table based on the representative chains selected from the latest release of PDB by our automatic construction system for ``PDB-REPRDB''. The development of the SSThread and the reconstruction of the 3D profile table have improved the prediction accuracy by 1% on average, respectively.
We present the new SSThread taking amino acid sequence homology at the structural aligned positions into account and the results of several prediction tests.

(1) Real World Computing Partnership (2) National Institute of Genetics