Protein Modelling and Surface Folding by Limiting the Degrees of Freedom - Computational Modeling in Tissue Engineering - page 19

method calculates the gradient or steep-out slope to obtain an energy mini-
mum [ 69 , 75 ].
1.2 The Monte Carlo method uses computational algorithms that rely on repeated
random sampling techniques to find a global minimum. The Monte Carlo
method takes a sample of the averages on the confirmation path; however, it is
possible for the path to enter into local minima that are difficult to differentiate
[ 13 , 89 ].
1.3 Smith's Microfibril model calculates the conformation energy by finding the
difference between the random state and the final conformation. The differ-
ence between the two is the objective function [ 17 , 43 , 67 ].
1.4 Probabilistic techniques based on Bayesian inference have been developed by
Garnier et al. [ 30 ] as information theory. The method takes into account the
probability of each amino acid having a particular secondary structure, and
also considers the assumptions and probabilities of each structure contributing
to that of its neighbours. The method is roughly 65 % accurate and is dra-
matically more successful in predicting alpha helices than beta sheets, which it
frequently erroneously predicted as loops or disorganized regions [ 52 ].
1.5 Neural Networks use training sets of solved structures to identify common
sequence motifs associated with particular arrangements of secondary struc-
tures. These methods are over 70 % accurate in their predictions, although
beta strands are still often under-predicted due to the lack of three-dimensional
structural information that would allow assessment of hydrogen bonding
patterns that can promote formation of the extended conformation required for
the presence of a complete beta sheet [ 52 ].
1.6 Support Vector Machine (SVM) is a method which analyzes data and rec-
ognizes patterns that are used for classification. SVM takes a set of input data
and predicts, for each given input, which of two possible classes forms the
input. SVM has proven particularly useful for predicting the locations of turns,
which are difficult to identify with statistical methods data [ 60 ]. The SVM
requirement for relatively small training sets has also been cited as an
advantage to avoid over-fitting to existing structures. The major limitation for
machine learning techniques is the attempt to predict more fine-grained local
properties of proteins, such as multiple back bone dihedral angle in unassigned
regions [ 87 ].
1.7 Genetic Algorithm (GA) is population of strings, which encode candidate
solutions to optimize a problem to evolve a better solution. The fundamental
Genetic Algorithm problems approach, due protein modelling interactions, are
elaborate, hence Genetic Algorithm evaluations for complex problems are
often the most prohibitive. Genetic Algorithms do not scale well with com-
plexity, meaning that with increasing numbers of elements there is an expo-
nential increase in the search space size. In general, the Genetic Algorithm has
a tendency to optimize to local minima, and therefore becomes globally
limited [ 79 ].
1.8 Ab initio models have been used exclusively to model binding sites of metals
to proteins [ 80 ] or use statistical methods to find a secondary structures