Transcript 赵亚辉
Ensembel classifier for protein fold pattern recognition Zhao Yahui 14S051028 Contents 1 Abstract 2 Introduction 3 Materials and Method 4 Result and Discussion 5 Conclusions Abstract Prediction of protein folding patterns(27) is one level deeper than that of protein structural classes, and hence is much more complicated and difficult. An ensemble classifier:a set of basic classifiers with each trained in different parameter(6) systems The operation engine for the constituent individual classifiers was OET-KNN rule. Introduction Current difficulty: local minimum problem when doesn't have stuctureknown homologous protein in the existing database Here: resort to the taxonomic approach(based on the assumption that the number of protein folds is limited) Materials and Methods Dataset: training set:311 protein testing set:385 protein None of proteins in these datasets has >35% sequence identity to any other, and most of proteins in the testing dataset have <25% sequence identity with those in the training dataset Materials and Methods Features: amino aid composition(replaced) predicted secondary structure hydrophobicity van der Waals volume polarity polarizability improve: pseudo-amino aid composition(avoid completely ignoring the sequence-order effects) Materials and Methods PseAA adopt the alternate correlation function between hydrophobicity and hydrophilicity( to reflect the sequence-order effects). Suppose a protein P with a sequence of L amino acid residues: is the hydrophobicity of , is the 1 2 hydrophilicity of . H i , j , H i , j are the hydrophobicity and hydrophilicity correlation functions Materials and Methods i , i 1 are the correspond ith-tier correlation factors that reflect the sequence-order correlation between all the ith most contiguous residues Materials and Methods Materials and Methods Standard conversion of h1 ( Ri ), h2 ( Ri ) : remain unchanged if going through the same conversion procedure again Materials and Methods By fusing the 2 amphiphilic correlation factors into the classical amino acid composition, we have the following augmented discrete form to represent a protein sample P: where OET-KNN Algorithm optimized evidence-theoretic k-nearest neighbors ET-KNN: Based on: each neighbor of a pattern to be classified is considered as an item of evidence supporting certain hypotheses concerning the class membership of that pattern. ET-KNN Algorithm Let us consider a problem of classifying N entities into 27 classes (fold types), which can be formulated as The available information is assumed to consist of a training dataset ET-KNN Algorithm Suppose P is a query protein to be classified, and the set of its k-nearest neighbors is for any , the knowledge that belongs to class gft can be considered as a piece of evidence that increases our belief that P also belongs to ,this item of evidence can be formulated by OET-KNN Algorithm The belief function of P belonging to class is a combination of its k-nearest neighbors, and can be formulated as: A decision is made by assigning the query protein P to the class : Materials and Methods Input Feature selection: (20+2 )D PseAA 21D predicted secondary structure, [(21*5)+(20+2 )]D 21D hydrophobicity, 21D normalized van der Waals volume, 21D polarity 21D polarizability Doing so would introduce too many parameters into the input, thereby reducing the clustertolerant capacity and cross-validation success rate It is instructive to introduce a new concept called the cluster-tolerant capacity for the dataset studied here. If such a capacity is high for a dataset, then the removal of any entry from the dataset will not significantly change the original clustered picture (such as the distribution of the standard vectors for each subset,the spacial gaps between the boundaries of any two subsets, and the original attribution of the removed entry to a subset) conversely, if the clustered tolerant capacity is low for a dataset, the removal of some entry from it will have a significant impact on the original clustered picture. Materials and Methods Framework of ensemble classifier: Why choose Frame work: reduce the variance caused by the peculiarities of a single training set and hence be able to learn a more expressive concept in classification than a single classifier. Materials and Methods Thus the query protein P is predicted belonging to the fold type with Suppose the ensemble classifier C is expressed by which its score is the highest： Thus, the process of how the ensemble classifier C works by fusing the nine basic classifiers C(i) (i =1,2……9) can be formulated as follows. Suppose Result And Discussion Conclusions It is shown thru the present study that the ensemble classifier formed by fusing different input types, particularly different dimensions of pseudo-amino acid composition, is very promising for enhancing Add your company slogan