Lukasz Kurgan

CAIM discretization algorithm

The task of extracting knowledge from databases is quite often performed by machine learning algorithms. The majority of these algorithms can be applied only to data described by discrete numerical or nominal attributes (features). In the case of continuous attributes, there is a need for a discretization algorithm that transforms continuous attributes into discrete ones. We describe such an algorithm, called CAIM (class-attribute interdependence maximization), which is designed to work with supervised data. The goal of the CAIM algorithm is to maximize the class-attribute interdependence and to generate a (possibly) minimal number of discrete intervals. The algorithm does not require the user to predefine the number of intervals, as opposed to some other discretization algorithms. The tests performed using CAIM and six other state-of-the-art discretization algorithms show that discrete attributes generated by the CAIM algorithm almost always have the lowest number of intervals and the highest class-attribute interdependency. Two machine learning algorithms, the CLIP4 rule algorithm and the decision tree algorithm, are used to generate classification rules from data discretized by CAIM. For both the CLIP4 and decision tree algorithms, the accuracy of the generated rules is higher and the number of the rules is lower for data discretized using the CAIM algorithm when compared to data discretized using six other discretization algorithms. The highest classification accuracy was achieved for data sets discretized with the CAIM algorithm, as compared with the other six algorithms.

A survey of Knowledge Discovery and Data Mining process models

Knowledge Discovery and Data Mining is a very dynamic research and development area that is reaching maturity. As such, it requires stable and well-defined foundations, which are well understood and popularized throughout the community. This survey presents a historical overview, description and future directions concerning a standard for a Knowledge Discovery and Data Mining process model. It presents a motivation for use and a comprehensive comparison of several leading process models, and discusses their applications to both academic and industrial problems. The main goal of this review is the consolidation of the research in this area. The survey also proposes to enhance existing models by embedding other current standards to enable automation and interoperability of the entire process.

D2P2: database of disordered protein predictions

We present the Database of Disordered Protein Prediction (D2P2), available at http://d2p2.pro (including website source code). A battery of disorder predictors and their variants, VL-XT, VSL2b, PrDOS, PV2, Espritz and IUPred, were run on all protein sequences from 1765 complete proteomes (to be updated as more genomes are completed). Integrated with these results are all of the predicted (mostly structured) SCOP domains using the SUPERFAMILY predictor. These disorder/structure annotations together enable comparison of the disorder predictors with each other and examination of the overlap between disordered predictions and SCOP domains on a large scale. D2P2 will increase our understanding of the interplay between disorder and structure, the genomic distribution of disorder, and its evolutionary history. The parsed data are made available in a unified format for download as flat files or SQL tables either by genome, by predictor, or for the complete set. An interactive website provides a graphical view of each protein annotated with the SCOP domains and disordered regions from all predictors overlaid (or shown as a consensus). There are statistics and tools for browsing and comparing genomes and their disorder within the context of their position on the tree of life.

Knowledge discovery approach to automated cardiac SPECT diagnosis

The paper describes a computerized process of myocardial perfusion diagnosis from cardiac single proton emission computed tomography (SPECT) images using data mining and knowledge discovery approach. We use a six-step knowledge discovery process. A database consisting of 267 cleaned patient SPECT images (about 3000 2D images), accompanied by clinical information and physician interpretation was created first. Then, a new user-friendly algorithm for computerizing the diagnostic process was designed and implemented. SPECT images were processed to extract a set of features, and then explicit rules were generated, using inductive machine learning and heuristic approaches to mimic cardiologists diagnosis. The system is able to provide a set of computer diagnoses for cardiac SPECT studies, and can be used as a diagnostic tool by a cardiologist. The achieved results are encouraging because of the high correctness of diagnoses.

MoRFpred, a computational tool for sequence-based prediction and characterization of short disorder-to-order transitioning binding regions in proteins

Motivation: Molecular recognition features (MoRFs) are short binding regions located within longer intrinsically disordered regions that bind to protein partners via disorder-to-order transitions. MoRFs are implicated in important processes including signaling and regulation. However, only a limited number of experimentally validated MoRFs is known, which motivates development of computational methods that predict MoRFs from protein chains. Results: We introduce a new MoRF predictor, MoRFpred, which identifies all MoRF types (α, β, coil and complex). We develop a comprehensive dataset of annotated MoRFs to build and empirically compare our method. MoRFpred utilizes a novel design in which annotations generated by sequence alignment are fused with predictions generated by a Support Vector Machine (SVM), which uses a custom designed set of sequence-derived features. The features provide information about evolutionary profiles, selected physiochemical properties of amino acids, and predicted disorder, solvent accessibility and B-factors. Empirical evaluation on several datasets shows that MoRFpred outperforms related methods: α-MoRF-Pred that predicts α-MoRFs and ANCHOR which finds disordered regions that become ordered when bound to a globular partner. We show that our predicted (new) MoRF regions have non-random sequence similarity with native MoRFs. We use this observation along with the fact that predictions with higher probability are more accurate to identify putative MoRF regions. We also identify a few sequence-derived hallmarks of MoRFs. They are characterized by dips in the disorder predictions and higher hydrophobicity and stability when compared to adjacent (in the chain) residues. Availability: http://biomine.ece.ualberta.ca/MoRFpred/; http://biomine.ece.ualberta.ca/MoRFpred/Supplement.pdf Contact: lkurgan@ece.ualberta.ca Supplementary information: Supplementary data are available at Bioinformatics online.

Impact of imputation of missing values on classification error for discrete data

Numerous industrial and research databases include missing values. It is not uncommon to encounter databases that have up to a half of the entries missing, making it very difficult to mine them using data analysis methods that can work only with complete data. A common way of dealing with this problem is to impute (fill-in) the missing values. This paper evaluates how the choice of different imputation methods affects the performance of classifiers that are subsequently used with the imputed data. The experiments here focus on discrete data. This paper studies the effect of missing data imputation using five single imputation methods (a mean method, a Hot deck method, a Nai@?ve-Bayes method, and the latter two methods with a recently proposed imputation framework) and one multiple imputation method (a polytomous regression based method) on classification accuracy for six popular classifiers (RIPPER, C4.5, K-nearest-neighbor, support vector machine with polynomial and RBF kernels, and Nai@?ve-Bayes) on 15 datasets. This experimental study shows that imputation with the tested methods on average improves classification accuracy when compared to classification without imputation. Although the results show that there is no universally best imputation method, Nai@?ve-Bayes imputation is shown to give the best results for the RIPPER classifier for datasets with high amount (i.e., 40% and 50%) of missing data, polytomous regression imputation is shown to be the best for support vector machine classifier with polynomial kernel, and the application of the imputation framework is shown to be superior for the support vector machine with RBF kernel and K-nearest-neighbor. The analysis of the quality of the imputation with respect to varying amounts of missing data (i.e., between 5% and 50%) shows that all imputation methods, except for the mean imputation, improve classification error for data with more than 10% of missing data. Finally, some classifiers such as C4.5 and Nai@?ve-Bayes were found to be missing data resistant, i.e., they can produce accurate classification in the presence of missing data, while other classifiers such as K-nearest-neighbor, SVMs and RIPPER benefit from the imputation.

SPINE X: improving protein secondary structure prediction by multistep learning coupled with prediction of solvent accessible surface area and backbone torsion angles.

Accurate prediction of protein secondary structure is essential for accurate sequence alignment, three‐dimensional structure modeling, and function prediction. The accuracy of ab initio secondary structure prediction from sequence, however, has only increased from around 77 to 80% over the past decade. Here, we developed a multistep neural‐network algorithm by coupling secondary structure prediction with prediction of solvent accessibility and backbone torsion angles in an iterative manner. Our method called SPINE X was applied to a dataset of 2640 proteins (25% sequence identity cutoff) previously built for the first version of SPINE and achieved a 82.0% accuracy based on 10‐fold cross validation (Q3). Surpassing 81% accuracy by SPINE X is further confirmed by employing an independently built test dataset of 1833 protein chains, a recently built dataset of 1975 proteins and 117 CASP 9 targets (critical assessment of structure prediction techniques) with an accuracy of 81.3%, 82.3% and 81.8%, respectively. The prediction accuracy is further improved to 83.8% for the dataset of 2640 proteins if the DSSP assignment used above is replaced by a more consistent consensus secondary structure assignment method. Comparison to the popular PSIPRED and CASP‐winning structure‐prediction techniques is made. SPINE X predicts number of helices and sheets correctly for 21.0% of 1833 proteins, compared to 17.6% by PSIPRED. It further shows that SPINE X consistently makes more accurate prediction in helical residues (6%) without over prediction while PSIPRED makes more accurate prediction in coil residues (3–5%) and over predicts them by 7%. SPINE X Server and its training/test datasets are available at http://sparks.informatics.iupui.edu/

Prediction of structural classes for protein sequences and domains-Impact of prediction algorithms, sequence representation and homology, and test procedures on accuracy

This paper addresses computational prediction of protein structural classes. Although in recent years progress in this field was made, the main drawback of the published prediction methods is a limited scope of comparison procedures, which in same cases were also improperly performed. Two examples include using protein datasets of varying homology, which has significant impact on the prediction accuracy, and comparing methods in pairs using different datasets. Based on extensive experimental work, the main aim of this paper is to revisit and reevaluate state of the art in this field. To this end, this paper performs a first-of-its-kind comprehensive and multi-goal study, which includes investigation of eight prediction algorithms, three protein sequence representations, three datasets with different homologies and finally three test procedures. Quality of several previously unused prediction algorithms, newly proposed sequence representation, and a new-to-the-field testing procedure is evaluated. Several important conclusions and findings are made. First, the logistic regression classifier, which was not previously used, is shown to perform better than other prediction algorithms, and high quality of previously used support vector machines is confirmed. The results also show that the proposed new sequence representation improves accuracy of the high quality prediction algorithms, while it does not improve results of the lower quality classifiers. The study shows that commonly used jackknife test is computationally expensive, and therefore computationally less demanding 10-fold cross-validation procedure is proposed. The results show that there is no statistically significant difference between these two procedures. The experiments show that sequence homology has very significant impact on the prediction accuracy, i.e. using highly homologous datasets results in higher accuracies. Thus, results of several past studies that use homologous datasets should not be perceived as reliable. The best achieved prediction accuracy for low homology datasets is about 57% and confirms results reported by Wang and Yuan [How good is the prediction of protein structural class by the component-coupled method?. Proteins 2000;38:165-175]. For a highly homologous dataset instance based classification is shown to be better than the previously reported results. It achieved 97% prediction accuracy demonstrating that homology is a major factor that can result in the overestimated prediction accuracy.

Prediction of protein structural class using novel evolutionary collocation-based sequence representation.

Knowledge of structural classes is useful in understanding of folding patterns in proteins. Although existing structural class prediction methods applied virtually all state‐of‐the‐art classifiers, many of them use a relatively simple protein sequence representation that often includes amino acid (AA) composition. To this end, we propose a novel sequence representation that incorporates evolutionary information encoded using PSI‐BLAST profile‐based collocation of AA pairs. We used six benchmark datasets and five representative classifiers to quantify and compare the quality of the structural class prediction with the proposed representation. The best, classifier support vector machine achieved 61–96% accuracy on the six datasets. These predictions were comprehensively compared with a wide range of recently proposed methods for prediction of structural classes. Our comprehensive comparison shows superiority of the proposed representation, which results in error rate reductions that range between 14% and 26% when compared with predictions of the best‐performing, previously published classifiers on the considered datasets. The study also shows that, for the benchmark dataset that includes sequences characterized by low identity (i.e., 25%, 30%, and 40%), the prediction accuracies are 20–35% lower than for the other three datasets that include sequences with a higher degree of similarity. In conclusion, the proposed representation is shown to substantially improve the accuracy of the structural class prediction. A web server that implements the presented prediction method is freely available at http://biomine.ece.ualberta.ca/Structural_Class/SCEC.html.

Improved sequence-based prediction of disordered regions with multilayer fusion of multiple information sources

Motivation: Intrinsically disordered proteins play a crucial role in numerous regulatory processes. Their abundance and ubiquity combined with a relatively low quantity of their annotations motivate research toward the development of computational models that predict disordered regions from protein sequences. Although the prediction quality of these methods continues to rise, novel and improved predictors are urgently needed. Results: We propose a novel method, named MFDp (Multilayered Fusion-based Disorder predictor), that aims to improve over the current disorder predictors. MFDp is as an ensemble of 3 Support Vector Machines specialized for the prediction of short, long and generic disordered regions. It combines three complementary disorder predictors, sequence, sequence profiles, predicted secondary structure, solvent accessibility, backbone dihedral torsion angles, residue flexibility and B-factors. Our method utilizes a custom-designed set of features that are based on raw predictions and aggregated raw values and recognizes various types of disorder. The MFDp is compared at the residue level on two datasets against eight recent disorder predictors and top-performing methods from the most recent CASP8 experiment. In spite of using training chains with ≤25% similarity to the test sequences, our method consistently and significantly outperforms the other methods based on the MCC index. The MFDp outperforms modern disorder predictors for the binary disorder assignment and provides competitive real-valued predictions. The MFDps outputs are also shown to outperform the other methods in the identification of proteins with long disordered regions. Availability: http://biomine.ece.ualberta.ca/MFDp.html Supplementary information: Supplementary data are available at Bioinformatics online. Contact: lkurgan@ece.ualberta.ca