Murat Gök

An ensemble of k-nearest neighbours algorithm for detection of Parkinson's disease

Parkinsons disease is a disease of the central nervous system that leads to severe difficulties in motor functions. Developing computational tools for recognition of Parkinsons disease at the early stages is very desirable for alleviating the symptoms. In this paper, we developed a discriminative model based on a selected feature subset and applied several classifier algorithms in the context of disease detection. All classifier performances from the point of both stand-alone and rotation-forest ensemble approach were evaluated on a Parkinsons disease data-set according to a blind testing protocol. The new method compared to hitherto methods outperforms the state-of-the-art in terms of both predictions of accuracy (98.46%) and area under receiver operating characteristic curve (0.99) scores applying rotation-forest ensemble k-nearest neighbour classifier algorithm.

A new feature encoding scheme for HIV-1 protease cleavage site prediction

HIV-1 protease has been the subject of intense research for deciphering HIV-1 virus replication process for decades. Knowledge of the substrate specificity of HIV-1 protease will enlighten the way of development of HIV-1 protease inhibitors. In the prediction of HIV-1 protease cleavage site techniques, various feature encoding techniques and machine learning algorithms have been used frequently. In this paper, a new feature amino acid encoding scheme is proposed to predict HIV-1 protease cleavage sites. In the proposed method, we combined orthonormal encoding and Taylor’s venn-diagram. We used linear support vector machines as the classifier in the tests. We also analyzed our technique by comparing some feature encoding techniques. The tests are carried out on PR-1625 and PR-3261 datasets. Experimental results show that our amino acid encoding technique leads to better classification performance than other encoding techniques on a standalone classifier.

Prediction of MHC class I binding peptides with a new feature encoding technique

The recognition of specific peptides, bound to major histocompatibility complex (MHC) class I molecules, is of particular importance to the robust identification of T-cell epitopes and thus the successful design of protein-based vaccines. Here, we present a new feature amino acid encoding technique termed OEDICHO to predict MHC class I/peptide complexes. In the proposed method, we have combined orthonormal encoding (OE) and the binary representation of selected 10 best physicochemical properties of amino acids derived from Amino Acid Index Database (AAindex). We also have compared our method to current feature encoding techniques. The tests have been carried out on comparatively large Human Leukocyte Antigen (HLA)-A and HLA-B allele peptide binding datasets. Empirical results show that our amino acid encoding scheme leads to better classification performance on a standalone classifier.

An Intelligent Power Factor Correction Approach Based on Linear Regression and Ridge Regression Methods

This study introduces an intelligent power factor correction approach based on Linear Regression (LR) and Ridge Regression (RR) methods. The 10-fold Cross Validation (CV) test protocol has been used to evaluate the performance. The best test performance has been obtained from the LR in comparison with RR. The empirical results have evaluated that the selected intelligent compensators developed in this work might overcome the problems met in the literature providing accurate, simple and low-cost solution for compensation.

A novel machine learning model to predict autism spectrum disorders risk gene

Autism spectrum disorders (ASD) are a group of neurodevelopmental disorders characterized by the difficulties of social interaction and communication skills, limited repetitive interests and behaviors. Diagnosis of ASD at the early stages is very desirable to establish the treatment quickly. From a machine learning view, the ASD risk prediction task is a binary classification of ASD risk genes that is whether a long non-coding RNA (lncRNA) gene causes disease or not. We have developed a machine learning model, trained using brain developmental gene expression data, for the binary classification of ASD risk genes. Our model is composed of two main parts: feature extraction with Haar wavelet transform, discretization methods and classification with Bayes network learning algorithm. We compared the proposed model with various standalone classifier algorithms and hitherto methods on lncRNA gene data. The experimental results confirm the efficiency of our model with sensitivity, area under ROC curve, MCC and F-measure scores of 0.902, 0.839, 0.583 and 0.806, respectively.

PROSES: A Web Server for Sequence-Based Protein Encoding

Recently, the number of the amino acid sequences shared in online databases is growing rapidly in huge amounts. By using sequence-derived features, machine learning algorithms are successfully applied to prediction of protein functional classes, protein-protein interactions, subcellular location, and peptides of specific properties in many studies. Protein Sequence Encoding System (PROSES) is a web server designed as freely and easily accessible for all researchers who want to use computational methods on protein sequence data. That is, PROSES provides users to encode their protein sequences easily without writing any programming code.

Prediction of Disordered Regions in Proteins Using Physicochemical Properties of Amino Acids

Disordered regions of proteins are highly abundant in various biological processes, involving regulation and signaling and also in relation with cancer, cardiovascular, autoimmune diseases and neurodegenerative disorders. Hence, recognizing disordered regions in proteins is a critical task. In this paper, we presented a new feature encoding technique built from physicochemical properties of residues selected as per the chaotic structure of related protein sequence. Our feature vector has been tested with various classification algorithms on an up-to-date data set and also compared to other methods. The proposed method shows better classification performance than many methods in terms of accuracy, sensitivity and specificity. Our results suggest that the new method that links the residues and their physicochemical properties using Lyapunov exponents is highly effective in recognition of disordered regions.