A Comparative Study of Methodologies of Protein Secondary Structure - Computational Intelligence Techniques for Comparative Genomics - page 39

E = extended strand in parallel and/or anti-parallel ʲ -sheet conformation. Min

length 2 residues.
B = residue in isolated

-sheet hydrogen bond formation)
S = bend (the only non-hydrogen-bond-based assignment)

ʲ -bridge (single pair ʲ

DSSP obtained from non-redundant PDB_select dataset, secondary structure
assigned by DSSP into eight conformational states to three states H = HGI, E = EB,
and C = STC.
GOR: The GOR method was rst developed by Garnier, Osguthrope, and

Robson [ 4 ]. It is an information theory-based method that uses a more powerful
probabilistic technique of Bayesian inference. This method not only takes into
account the probability of each amino acid having a particular secondary structure
but also the conditional probability of the amino acid assuming each structure given
the contributions of its neighbors. This method assumes that amino acids up to eight
residues on each side that in uence the secondary structure of the central residue.

This program has many versions but here we are going to write the program in the
fourth version. The algorithm makes use a sliding window of 17 amino acids. All
possible pairs of amino acids in this window are checked for their information
content as to predict the central amino acid by comparing the set of 266 other
proteins of known structure. In this version, we are going to

fl

nd the certain pair-
wise combinations of amino acids that are present in the region called as

anking
region. If a particular amino acid is surrounded by residues that prefer to be a helix,
it is likely to be a helix, even if its individual helical preference is low. This method
helps in considering the propensity of a single residue and calculates the position
dependent propensities for helics, sheet and turn and thus simultaneously for all
types of residues.
GOR V: This method is a bit improved method when compared to GOR IV
method [ 5 ]. The improvement was made in the algorithm. It was inclusion of
evolutionary information using PSI-BLASTMultiple alignments are generated
using PSI-BLAST after

fl

ve iterations based on the non-redundant database. The
GOR V method prediction takes a longer time for the multiple alignments that is
this kind of alignments takes much more time for hits.
Accuracy of the two methods: The prediction accuracy of the GOR IV method
is 64.4 % and the prediction accuracy of GOR V method is 73.5 %. Even though
GOR V method has more prediction among accuracy, people generally use the
GOR IV method. This means that GOR V method is dif cult to implement.