Evolutionary Granular Kernel Trees and Applications in Drug Activity Comparisons

With the growing interest of biological data prediction and chemical data prediction, more complicated kernels are designed to measure data similarities. In (1), we proposed a kind of evolutionary granular kernel trees (EGKTs) for drug activity comparisons. In EGKTs, feature granules and tree structures are predefined based on the possible substituent locations. In (2), we proposed a granular kernel tree structure evolving system (GKTSES) to evolve the structures of GKTs in the case that we lack knowledge to predefine kernel trees. In this paper, evolutionary voting kernel machines (EVKMs) are presented based on GKTSES. Experimental results show that EVKMs are more stable than GKTSES in cyclooxygenase-2 inhibitor activity comparisons. Index Terms—Drug activity comparisons, kernel, genetic algorithms, granular kernel trees, support vector machines.

Some sets of orthogonal polynomial kernel functions

A novel semantic smoothing kernel for text classification with class-based weighting

A procedure that generates random conformations of a protein chain, and then applies energy minimization to find the structure of lowest energy, is described. Single‐residue conformations are represented in terms of four conformational states, α, ϵ, α*, and ϵ*. Each state corresponds to a rectangular region in the ϕ, ψ map. The conformation of an entire chain is then represented by a sequence of single‐residue conformational states. The distinct “chain‐states” in this representation correspond to multidimensional rectangular regions in the conformational space of the whole protein. A set of highly‐probable chain‐states can be predicted from the amino acid sequence using the pattern recognition procedure developed in the first two articles of this series. The importance‐sampling minimization procedure of the present article is then used to explore the regions of conformational space corresponding to each of these chain‐states. The importance‐sampling procedure generates a number of random conformations within a particular multidimensional rectangular region, sampling most densely from the most probable, or “important,” sections of the ϕ, ψ map. All values of ϕ and ψ are allowed, but the less‐probable values are sampled less often. To achieve this, the random values of ϕ and Φ are generated from bivariate gaussian distributions that are determined from known X‐ray structures. Separate gaussian distributions are used for proline residues in the α and ϵ states, for glycine residues in the α, ϵ, α*, and ϵ* states, and for ordinary residues involved in 29 different tripeptide conformations. Energy minimization is then applied to the randomly‐generated structures to optimize interactions and to improve packing. The final energy values are used to select the best structures. The importance‐sampling minimization procedure is tested on the avian pancreatic polypeptide, using chain‐states predicted from the amino acid sequence. The conformation having the lowest energy is very similar to the X‐ray conformation.

Advancement of science and technology has prompted researchers to develop new intelligent systems that can solve a variety of problems such as pattern recognition, prediction, and optimization. The ability of the human brain to learn in a fashion that tolerates noise and error has attracted many researchers and provided the starting point for the development of artificial neural networks: the intelligent systems. Intelligent systems can acclimatize to the environment or data and can maximize the chances of success or improve the efficiency of a search. Due to massive parallelism with large numbers of interconnected processers and their ability to learn from the data, neural networks can solve a variety of challenging computational problems. Neural networks have the ability to derive meaning from complicated and imprecise data; they are used in detecting patterns, and trends that are too complex for humans, or other computer systems. Solutions to the toughest problems will not be found through one narrow specialization; therefore we need to combine interdisciplinary approaches to discover the solutions to a variety of problems. Many researchers in different disciplines such as medicine, bioinformatics, molecular biology, and pharmacology have successfully applied artificial neural networks. This chapter helps the reader in understanding the basics of artificial neural networks, their applications, and methodology; it also outlines the network learning process and architecture. We present a brief outline of the application of neural networks to medical diagnosis, drug discovery, gene identification, and protein structure prediction. We conclude with a summary of the results from our study on tuberculosis data using neural networks, in diagnosing active tuberculosis, and predicting chronic vs. infiltrative forms of tuberculosis.

This paper is an attempt to highlight the fundamental approaches, algorithms and techniques being employed in interesting areas of Computational Biology. The comparative analysis of genomic sequences, micro-array technology, phylo-genetic tree, alignment matters, maximum sequence search and protein structure prediction are some of NP hard problems. There has been no direct solution proposed so far. Approximate and optimal solutions have been devised using different tools. We have compared different approaches in various domains with the aid of exemplary data. We assigned different weight values for relevant features and obtained scores in comparison and alignment phenomenon. We classified sequence analysis to be a broader domain encapsulating visualization, pattern recognition, motif concentration and structure prediction. Different techniques and approaches have been evaluated and compared. It was analyzed that some approaches suitable for certain domains may not be recommended to employ for other domains. The usefulness of this analysis has been explored and further future expectations have been described.

Matching two geometric objects in two-dimensional (2D) and three-dimensional (3D) spaces is a central problem in computer vision, pattern recognition, and protein structure prediction. In particular, the problem of aligning two polygonal chains under translation and rotation to minimize their distance has been studied using various distance measures. It is well known that the Hausdorff distance is useful for matching two point sets, and that the Fréchet distance is a superior measure for matching two polygonal chains. The discrete Fréchet distance closely approximates the (continuous) Fréchet distance, and is a natural measure for the geometric similarity of the folded 3D structures of biomolecules such as proteins. In this paper, we present new algorithms for matching two polygonal chains in two dimensions to minimize their discrete Fréchet distance under translation and rotation, and an effective heuristic for matching two polygonal chains in three dimensions. We also describe our empirical results on the application of the discrete Fréchet distance to protein structure-structure alignment.

ML has revolutionised bioinformatics' difficult biological data analysis. Pattern recognition and biological process categorization help ML systems diagnose diseases, predict protein structures, and investigate gene expression. Few-shot learning and bioinformatics are effective at optimising results with limited resources, overcoming biological dataset access issues. ML and bioinformatics advance precision medicine and drug discovery while improving biological understanding. This chapter examines bioinformatics ML methods like supervised classification, clustering, and probabilistic graphical models to find new insights. Text mining, systems biology, evolution, proteomics, and genomics use deterministic and stochastic heuristics for optimisation. Understanding bioinformatics methods and modern ML technologies while understanding implementation challenges is the study goal. Few-shot learning is highlighted to show its importance. ML and bioinformatics together improve our knowledge and solve real problems, improving research methods in biological and medical sciences.

The qualities of clustering, including those obtained by the kernel-based methods should be assessed. In this paper, by investigating the inherent pairwise similarities in kernel matrix implicitly defined by the kernel function, we define two statistical similarity coefficients which can be used to describe the within-cluster and between-cluster similarities between the data items, respectively. And then, an efficient cluster validity index and a self-adaptive kernel clustering (SAKC) algorithm are proposed based on these two similarity coefficients. The performance and effectiveness of the proposed validity index and SAKC algorithm are demonstrated, compared with some existing methods, on two synthetic datasets and four UCI real databases. And the robustness of this new index with Gaussian kernel width is also explored tentatively.KeywordsData ItemSynthetic DatasetKernel MatrixValidity IndexRing DataThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Feature extraction using fuzzy maximum margin criterion

Sprinkled semantic diffusion kernel for word sense disambiguation

A framework of spectrum sensing with a multi-class hypothesis is proposed to maximize the achievable throughput in cognitive radio networks. The energy range of a sensing signal under the hypothesis that the primary user is absent (in a conventional two-class hypothesis) is further divided into quantized regions, whereas the hypothesis that the primary user is present is conserved. The non-radio frequency energy harvesting-equiped secondary user transmits, when the primary user is absent, with transmission power based on the hypothesis result (the energy level of the sensed signal) and the residual energy in the battery: the lower the energy of the received signal, the higher the transmission power, and vice versa. Conversely, the lower is the residual energy in the node, the lower is the transmission power. This technique increases the throughput of a secondary link by providing a higher number of transmission events, compared to the conventional two-class hypothesis. Furthermore, transmission with low power for higher energy levels in the sensed signal reduces the probability of interference with primary users if, for instance, detection was missed. The familiar machine learning algorithm known as a support vector machine (SVM) is used in a one-versus-rest approach to classify the input signal into predefined classes. The input signal to the SVM is composed of three statistical features extracted from the sensed signal and a number ranging from 0 to 100 representing the percentage of residual energy in the node’s battery. To increase the generalization of the classifier, k-fold cross-validation is utilized in the training phase. The experimental results show that an SVM with the given features performs satisfactorily for all kernels, but an SVM with a polynomial kernel outperforms linear and radial-basis function kernels in terms of accuracy. Furthermore, the proposed multi-class hypothesis achieves higher throughput compared to the conventional two-class hypothesis for spectrum sensing in cognitive radio networks.

Animal models that recapitulate human epilepsies with more or less details are believed to be of importance for clinics – for studies of anticonvulsants (Reigel et al., 1986, Dailey et al., 1996, Fedotova et al., 1996, Kosacheva et al., 1998, Ross, Coleman, 2000 et al.), for dissection of molecular and biochemical pathogenesis of epilepsy, and for the search of epilepsy susceptibility genes. The use of audiogenic epilepsy in rodents as the model for anticovulsant activity of different drugs is popular, the PubMed search for the key words combination “anticonvulsant and audiogenic and model” resulted in about 100 citations. Thus the detailed investigation of this type of seizures has great value for iur knowledge concerning the genesis of seizure (and epileptogenesis) in general and their genetic basis in particular. Apart of its practical importance, the rodent audiogenic epilepsy is the enigmatic biological and genetic phenomenon. Actually the biological significance of high sound sensitivity for rodent survival was never questioned but at the same time its connection with audiogenic epilepsy (AE) was not analyzed from this point of view either. This chapter aims first to introduce several data items concerning investigations of Russian audiogenic prone rat strain (Krushinsky-Molodkina, KM) as they were not fully represented in English literature and second – to discuss the general biological mechanisms of audiogenic seizure. It is well known that the good theory is the best friend of practice. There is some hope that elucidating the origin of animal audiogenic epilepsies will bring us closer to still unsolved problems of seizure states in general. The methods which were used in studies of audiogenic epilepsies marched along with neurobiology methods starting from the middle of XX century. The same is now – new technologies are going to be applied and will be mobilized in future as these valuable models (seizure states developing in response to loud sound) are of great use for medical practice and clinic in particular.

K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) are the two Supervised Machine Learning Algorithms which have been used extensively in solving classification and regression problems in various application domains—medical diagnosis, pattern recognition, and image classification being the popular ones. KNN has been attractive for its simplicity—SVM has been considered as complex. Computational complexity of KNN is less compared to SVM. However, accuracy of SVM in general has been higher compared to KNN. This paper carries out an analysis of complexity of KNN and SVM and then looks at the alternative strategies to reduce the complexity. Accuracies of both SVM and KNN are analyzed from literature survey of experimental results with different training datasets. Reduction of number of training instances and usage of kernel functions are found to be the solutions for reducing SVM complexity. Factors impacting KNN accuracy are analyzed and possible alternatives are assessed. Kernel functions and distance-weighted KNN are concluded as the solution to improve KNN accuracy to compete with SVM accuracy. To utilize the best potential of the two algorithms, the option of a hybrid algorithm combining both KNN and SVM is also proposed.

One-against-rest (OAR) is a well known multiclassification structure, which is an extension from binary classifiers. It has shown its great potential in pattern recognition and hyperspectral data processing. However, existence of unclassified region limits its application. In this paper, a new multiclassifier based on OAR combined with one-against-one (OAO) structure is proposed. In the multiclassifier, OAO is used to classify the unclassified region to improve performance of OAR. At the same time, the formation of unclassified region is discussed, and the pattern of selecting classifiers for secondary classification on unclassified region is proposed. To compare secondary classifiers and prove the conclusion, other six classifiers are selected , which are decision tree (DT), minimum distance (MD) based on Euclidean distance, MD based on Euclidean distance with kernel function, MD based on Mahalanobis distance, spectral mapping classifier (SMC) and maximum likelihood classifier (MLC). The SVM is used for OAR, OAO and DT in experiment and a hyperspectral remote sensing image is used as testing samples.