WANG Fangping et al: Prediction of Protein Structural Classes Using … 261
sition (PseAAC). The PseAAC not only contains AAC but also incorporates the information of sequence order. Based on the concept of the PseAAC, Chou and Liu originally selected 400 dipeptide compositions as in-formation parameters to predict protein structural classes by the theory of conditional probability and im-proved prediction accurate rate of protein secondary structure[14].
The measure of diversity (MD) is a kind of infor-mation description on state space[15]. On the basis of the theory of MD, Laxton introduced the theory of incre-ment of diversity (ID)[15], which can quantificationally describe similarity level of two datasets. The ID theory was applied for the prediction of protein structural classes by Li et al[16], in which the number of α-helices, β-sheets, and βαβ mixed fragments are used as informa-tion parameters. Besides, according to recent results by Lin and Li[17], using AAC and 400 dipeptide composi-tions, respectively, the total predictive success rate of the latter is higher than that of the former by increment of diversity (ID) algorithm. Several SVM-based meth-ods have been developed for predicting protein domain structural classes[9-12]. Compared with other machine learning systems, the SVM has many attractive features, including high speed, excellent classification perform-ance, and capability of condensing information con-tained in the training set[18]. In this article, the 400 dipeptide compositions were selected as parameters of source of diversity, and the theory of ID was used to predict protein structural classes. For the purpose of improving the predictive success rate, the SVM carried out classification again.
D(X)=D(n1,n2,",ns)=NlnN ∑nilnni (1)
i=1
s
Here, N=∑ni.
i=1
s
In the same of state space, if there are two sources of diversity:X:[n1,n2,",ns]andY:[m1,m2,",ms], the ID can be defined as
ID(X,Y)=D(X+Y) D(X) D(Y) (2)
1.2.2 Calculation of the ID value using given parame-ters as diversity source
In this study, the absolute occurrence frequencies of 400 dipeptides are selected as the information parameters of a diversity source X400=[N1,N2,",N400]. Therefore, using Eq. (1), for any protein X, the measure of diversity can be expressed as
D(X)=NlnN ∑NilnNi (3)
i=1
400
where N=∑Ni.
i=1
400
The four standard sources of diversity for all-α, all-β, α/β, and α+β classes can be expressed as
ξ
Xξ:[N1ξ,N2,",Nsξ],(ξ=α,β,α/β,α+β)
where Xξ represents the numbers of information pa-rameters of the ξ class proteins.
The four standard measures of diversity can be ex-pressed as
D(Xξ)=NξlnNξ ∑NiξlnNiξ,
i=1400
(ξ=α,β,α/β,α+β) (4)
where Nξ=∑Niξ.
i=1400
1 Materials and Methods
1.1 Dataset
For convenience of comparison, the dataset of 204 proteins constructed by Chou are used as the predictive dataset (doc.guandang.net), which can be classified into four structural classes: 52 all-α, 61 all-β, 45 α/β, and 46 α+β according to the definition by Levitt and Chothia[1]. 1.2 Methods
1.2.1 The definition of increment of diversity
For a state space X:{x1,x2,",xs}consisting of s
information symbols, if ni indicates the number of the i th state, the measure of diversity for diversity source X:[n1,n2,",ni,",ns] is defined as[16]
Given a protein X whose protein structural class is to be
predicted, its measure of diversity can be expressed as Eq. (3); then, using Eq. (2), the ID between the sources of div-
ξξ
ersityXξ:[N1ξ,N2,",N400] and X:[N1,N2,",N400] is calculated by the following formula:
ID(Xξ,X)=D(Xξ+X) D(Xξ) D(X),
(ξ=α,β,α/β,α+β) (5)
here, D(Xξ+X) denotes the measure of diversity source: Xξ +X calculated by Eq.(1).
Therefore, for any protein X to be predicted, one has four IDs,(ID(Xα,X),ID(Xβ,X),ID(Xα+β,X), and ID (Xα/β,X)), corresponding to four standard measures of diversity.
1.2.3 ID-SVM approach predict protein structural class
In this research, the publicly available LIBSVM
Prediction of Protein Se... 暂无评价 9页 免费...proteins using the Internet resources are described...structural diversity of proteins begins with ...
Prediction of protein protein interaction types using the decisionPrediction of protein protein interaction types using the decision隐藏>> Mathematical and ...
Prediction of Protein Topologies Using GIOHMMs and GRNNs Department of Information and Computer Science University of California, Irvine Irvine, CA 92697-3425...
Prediction of the secondary structural content might be considered the ?rst step in analyzing a new protein sequence and predicting its tertiary architecture ...
Structural Prediction of Membrane Protein: Application to Known Structures_专业资料。Joma fDogu iest(nulonhaUnvriyEg.E)V12d.o.7,No.(00)221 2 89 ...
feature selection using markov blanket induction for the prediction of protein...351 molecular structural properties that predict binding to thrombin (and thus...
Using the discovered associative feature we predict implicit protein interactions...theory based feature dimension reduction filter with fea- Prediction of ...
12页 1财富值 Prediction of Protein St... 暂无评价 5页 5财富值...更多>>We show that the rst-order theory of structural subtyping of non-...
12页 1财富值 Prediction of Protein St... 暂无评价 5页 5财富值... On the Theory of Structural Subtyping Viktor Kuncak and Martin Rinard ...
Sharan et al Box 2 Function prediction using the MRF method The Markov ...Data of a high diversity and dimensionality have been integrated using ...
热门文档