Jayaraman Valadi

Applications of Metaheuristics in Process Engineering

Metaheuristics exhibit desirable properties like simplicity, easy parallelizability and ready applicability to different types of optimization problems such as real parameter optimization, combinatorial optimization and mixed integer optimization. They are thus beginning to play a key role in different industrially important process engineering applications, among them the synthesis of heat and mass exchange equipment, synthesis of distillation columns and static and dynamic optimization of chemical and bioreactors. This book explains cutting-edge research techniques in related computational intelligence domains and their applications in real-world process engineering. It will be of interest to industrial practitioners and research academics.

Parallel Ant-Miner (PAM) on High Performance Clusters

This study implements parallelization of Ant-Miner for classification rules discovery. Ant-Miner code is parallelized and optimized in a cluster environment by employing master-slave model. The parallelization is achieved in two different operations of Ant-Miner viz. discretization of continuous attributes and rule construction by ants. For rule mining operation, ants are equally distributed into groups and sent across the different cluster nodes. The performance study of Parallel Ant-Miner (PAM) employs different publicly available datasets. The results indicate remarkable improvement in computational time without compromising on the classification accuracy and quality of discovered rules. Dermatology data having 33 features and musk data having 168 features were taken to study performance with respect to timings. Speedup almost equivalent to ideal speedup was obtained on 8 CPUs with increase in number of features and number of ants. Also performance with respect to accuracies was done using lung cancer data.

Recent trends in antimicrobial peptide prediction using machine learning techniques

The importance to develop effective alternatives to known antibiotics due to increased microbial resistance is gaining momentum in recent years. Therefore, it is of interest to predict, design and computationally model Antimicrobial Peptides (AMPs). AMPs are oligopeptides with varying size (from 5 to over100 residues) having key role in innate immunity. Thus, the potential exploitation of AMPs as novel therapeutic agents is evident. They act by causing cell death either by disrupting the microbial membrane by inhibiting extracellular polymer synthesis or by altering intra cellular polymer functions. AMPs have broad spectrum activity and act as first line of defense against all types of microorganisms including viruses, bacteria, parasites, fungi and as well as cancer (uncontrolled celldivision) progression. Large-scale identification and extraction of AMPs is often non-trivial, expensive and time consuming. Hence, there is a need to develop models to predict AMPs as therapeutics. We document recent trends and advancement in the prediction of AMP.

Classification of anti hepatitis peptides using Support Vector Machine with hybrid Ant Colony OptimizationThe Luxembourg database of trichothecene type B F. graminearum and F. culmorum producers.

Hepatitis is an emerging global threat to public health due to associated mortality, morbidity, cancer and HIV co-infection. Available diagnostics and therapeutics are inadequate to intercept the course and transmission of the disease. Antimicrobial peptides (AMP) are widely studied and broad-spectrum host defense peptides are investigated as a targeted anti-viral. Therefore, it is of interest to describe the supervised identification of anti-hepatitis peptides. We used a hybrid Support Vector Machine (SVM) with Ant Colony Optimization (ACO) algorithm for simultaneous classification and domain feature selection. The described model shows a 10 fold cross-validation accuracy of 94 percent. This is a reliable and a useful tool for the prediction and identification of hepatitis specific drug activity

Hybrid ACO Chaos-Assisted Support Vector Machines for Classification of Medical Datasets

There is a need for developing accurate learning algorithms for analyzing large-scale medical diagnostic, prognostic, and treatment datasets. Success of classifiers like support vector machines lies in employment of best informative features out of a huge noisy feature space. In this work, we employ a hybrid filter–wrapper approach to build high-performance classification models. We tested our algorithms using popular datasets containing clinic-bio-pathological parameters of leukemia, hepatitis, breast cancer, and colon cancer taken from publically available datasets. Our results indicate that the hybrid algorithm can discover informative subsets possessing very high classification accuracy.

Glowworm Swarm Based Informative Attribute Selection Using Support Vector Machines for Simultaneous Feature Selection and Classification

In this paper, we propose a hybrid filter-wrapper algorithm, GSO-Infogain, for simultaneous feature selection for improved classification accuracy. GSO-Infogain employs Glowworm-Swarm Optimization(GSO) algorithm with Support Vector Machine(SVM) as its internal learning algorithm and utilizes feature ranking based on information gain as a heuristic. The GSO algorithm randomly generates a population of worms, each of which is a candidate subset of features. The fitness of each candidate solution, which is evaluated using Support Vector Machine, is encoded within its luciferin value. Each worm probabilistically moves towards the worm with the highest luciferin value in its neighbourhood. In the process, they explore the feature space and eventually converge to the global optimum. We have evaluated the performance of the hybrid algorithm for feature selection on a set of cancer datasets. We obtain a classification accuracy in the range 94-98 % for these datasets, which is comparable to the best results from other classification algorithms. We further tested the robustness of GSO-Infogain by evaluating its performance on the CoEPrA training and test datasets. GSO-Infogain performs well in this experiment too by giving similar prediction accuracies on the training and test datasets thus indicating its robustness.

Simultaneous Gene Selection and Cancer Classification Using a Hybrid Intelligent Water Drop Approach

Computational Analysis of gene expression data is extremely difficult, due to the existence of a huge number of genes and less number of samples (limited number of patients). Thus,it is of significant importance to provide a subset of the most informative genesto a learning algorithm, for constructing robust prediction models. In this study, we propose a hybrid Intelligent Water Drop (IWD) - Support Vector Machines (SVM) algorithm, with weighted gene ranking as a heuristic, for simultaneous gene subset selection and cancer prediction. Our results, evaluated on three cancer datasets, demonstrate that the genes selected by the IWD technique yield classification accuracies comparable to previously reported algorithms.

Simultaneous gene selection and cancer classification using a hybrid group search optimizer

Constructing classifier models for gene expression datasets using informative features enhances prediction performance of concerned models. Here, we propose a hybrid Group Search based feature selection (GSO-FS) algorithm which can select relevant gene subsets that can optimally predict cancerous tissue samples. Our experimental results show that the GSO-FS algorithm in combination with SVM classifier performs quite well.

Hybrid feature selection and peptide binding affinity prediction using an EDA based algorithm

Protein function prediction is an important problem in functional genomics. Typically, protein sequences are represented by feature vectors. A major problem of protein datasets that increase the complexity of classification models is their large number of features. The process of drug discovery often involves the use of quantitative structure-activity relationship (QSAR) models to identify chemical structures that could have good inhibitory effects on specific targets and have low toxicity (non-specific activity). QSAR models are regression or classification models used in the chemical and biological sciences. Because of high dimensionality problems, a feature selection problem is imminent. In this study, we thus employ a hybrid Estimation of Distribution Algorithm (EDA) based filter-wrapper methodology to simultaneously extract informative feature subsets and build robust QSAR models. The performance of the algorithm was tested on the benchmark classification challenge datasets obtained from the CoePRa competition platform, developed in 2006. Our results clearly demonstrate the efficacy of a hybrid EDA filter-wrapper algorithm in comparison to the results reported earlier.

Twelve years of BIOinformatics CLUb for Experimenting Scientists (Bioclues)

We describe the scientific community of an India effort, the Bioinfomatics Club for Experimenting Scientists (Bioclues) whose aim works on four avenues, viz. Mentoring, Outreach, Research and Entrepreneurship (MORE). Incepted in the year 2005, the organization went on to become an affiliate of International Society for Computational Biology (ISCB) in 2011. Ably supported by Asia pacific Bioinformatics Network (APBioNet), we are one of the fastest growing bioinformatics societies in India, currently serving over 3400 members from nearly 30 countries. Bioclues adheres Creative Commons License with the prime focus to help the bioinformaticists in India to promote open access. In the year 2010, when we setup vision 2020, we aimed to bring together the Indian bioinformaticians, foster a strong working mentor-mentee relationship, provide access to bioinformatics resources, organize conferences and workshops besides imparting information about research, training, education, employment and current events and news from bioinformatics, genomics, and related fields. In this article, we describe the challenges across the four avenues and further highlight the opportunities the organization has met the last decade in understanding the core necessity of computational biology virtual projects driven by these avenues viz. MORE.