Secondary and Tertiary Structure Prediction of Proteins: A Bioinformatic Approach - Complex System Modelling and Control Through Intelligent Soft Computations - page 534

Table 1 Current PDB holdings (as on April 22nd, 2014)
Experimental methods Molecule types

Proteins Nucleic

acids

Protein/NA
complexes

Other Total

X ray 82,406 1,516 4,287 4 88,213

NMR 9,129 1,078 206 7 10,420

Electron microscopy 523 52 173 0 748

Hybrid 59 3 2 1 65

Other 155 4 6 13 178

Total 92,272 2,653 4,674 25 99,624

highly needed to fill the gap between the protein sequences available in public

domain databases and their experimentally solved structures.
Historically, protein structure prediction was classi

ed into three categories: (i)
Ab initio modeling (Liwo et al. 1999 ; Zhang et al. 2003 ; Bradley et al. 2005 ;
Klepeis et al. 2005 ; Klepeis and Floudas 2003 ) (ii) Threading or Fold recognition
(Bowie et al. 1991 ; Jones et al. 1992 ; Xu and Xu 2000 ; Zhou and Zhou 2005 ;
Skolnick et al. 2004 ) and (iii) Homology or Comparative modeling ( ali and

Blundell 1993 ; Fiser et al. 2000 ). Threading and comparative modeling build
protein models by aligning query sequences onto solved template structures by X-
ray crystallography or NMR. When close templates are identi

Š

ed, high-resolution
models could be built by the template-based methods. If templates are absent from
the PDB, the models need to be built from scratch, i.e. ab initio modeling.
Nowadays, these prediction categories are clubbed into two major groups: free
modeling (FM) involving Ab initio folding and template-based modeling (TBM),
which includes comparative/homology modeling and threading. These predicted
models must be checked for protein structure quality validation by various pro-
grammes available.
This chapter is broadly divided under 9 sections which are further divided into
sub-headings wherever required. Section 2.1 describes about amino acid propensity
based secondary structure prediction method. Section 2.2 discusses about template
based secondary structure predictions and the accuracy obtained by these methods.
Section 2.3 explains the secondary structure prediction methods based on machine
learning approaches. Ab initio folding/modeling and its limitations are is described
in Sect. 3.1 . Threading and Homology modeling methods with their strengths and
their weakness are explained in Sects. 3.2 and 3.3 respectively. Hybrid and Meta-
Servers which aid in accuracy of protein models are described in Sects. 4 and 5 .
Section 6 describes about the protein structure prediction community, Critical
Assessment of protein Structure Prediction (CASP). Section 7 describes about the
various application of protein models generated by the three major prediction
methods. Future prospects of protein secondary and tertiary structure prediction