Information Technology Reference
In-Depth Information
2.3 Sequence Pro
le Based Method
This method uses the machine learning algorithms to predict the secondary struc-
ture of the query protein. Arti
cial Neural Networks (ANNs), Support Vector
Machines (SVMs) and Hidden Markov Models (HMMs) are the most widely used
machine learning algorithms that come under this category (Jones 1999a ; Karplus
et al. 1998 ; Kim and Park 2003 ; Chandonia and Karplus 1995 ). Currently, most
effective PSS prediction methods are based on machine learning algorithms, such as
PSIPRED (McGuf
n et al. 2000 ), SVMpsi (Kim and Park 2003 ), PHD (Rost et al.
1994 ), PHDpsi (Przybylski and Rost 2002 ), Porter (Pollastri and McLysaght 2005 ),
JPRED3 (Cole et al. 2008 ), STRIDE (Heinig and Frishman 2004 ), SPARROW
(Bettella et al. 2012 ) and SOPMA (Geourjon and Del
é
age 1995 ) and which employ
Arti
cial Neural Network (ANN) or Support Vector Machines (SVM) learning
models. In addition to protein secondary structure, these servers also make pre-
dictions on Solvent Accessibility and Coiled-coil regions etc. These programmes or
web-servers are listed in Table 2 . These methods have an accuracy ranging
72
80 %, depending on the method, the training and the test datasets.
Two types of errors are most prevalent in secondary structure prediction of
proteins. One of these errors is called local errors which occur when a residue is
wrongly predicted. Second type of error is called structural error, which occur when
the structure is altered globally. Sometimes, errors that alter the function of a
protein should be avoided whenever possible. Q3 is the most commonly used
measures of local errors, whereas the Segment Overlap (SOV) Score (Zemla et al.
1999 ) is the most well known measure for structural errors. These measures have
been adopted by various communities in these research areas e.g. CASP (Moult
et al. 1995 ) and EVA (Eyrich et al. 2001 ). Good secondary structures lay the
foundation for better prediction of tertiary structures of proteins. The following
section provides an insight into the methods for predicting the tertiary structures of
proteins.
-
3 Tertiary Structure Prediction
As discussed in introduction, tertiary structure prediction methods are categorized
into three major methods to model a target protein sequence. Flowchart for
selecting the most accurate prediction algorithm/method among these three cate-
gories for the target sequence is schematically represented in Fig. 1 .
Search WWH ::




Custom Search