Yamuna Prasad

SVM classifier based feature selection using GA, ACO and PSO for siRNA design

Recently there has been considerable interest in applying evolutionary and natural computing techniques for analyzing large datasets with large number of features. In particular, efficacy prediction of siRNA has attracted a lot of researchers, because of large number of features involved. In the present work, we have applied the SVM based classifier along with PSO, ACO and GA on Huesken dataset of siRNA features as well as on two other wine and wdbc breast cancer gene benchmark dataset and achieved considerably high accuracy and the results have been presented. We have also highlighted the necessary data size for better accuracy in SVM for selected kernel. Both groups of features (sequential and thermodynamic) are important in the efficacy prediction of siRNA. The results of our study have been compared with other results available in the literature.

Protein-protein interaction and molecular dynamics analysis for identification of novel inhibitors in Burkholderia cepacia GG4

The lack of complete treatments and appearance of multiple drug-resistance strains of Burkholderia cepacia complex (Bcc) are causing an increased risk of lung infections in cystic fibrosis patients. Bcc infection is a big risk to human health and demands an urgent need to identify new therapeutics against these bacteria. Network biology has emerged as one of the prospective hope in identifying novel drug targets and hits. We have applied protein-protein interaction methodology to identify new drug-target candidates (orthologs) in Burkhloderia cepacia GG4, which is an important strain for studying the quorum-sensing phenomena. An evolutionary based ortholog mapping approach has been applied for generating the large scale protein-protein interactions in B. Cepacia. As a case study, one of the identified drug targets; GEM_3202, a NH (3)-dependent NAD synthetase protein has been studied and the potential ligand molecules were screened using the ZINC database. The three dimensional structure (NH (3)-dependent NAD synthetase protein) has been predicted from MODELLERv9.11 tool using multiple PDB templates such as 3DPI, 2PZ8 and 1NSY with sequence identity of 76%, 50% and 50% respectively. The structure has been validated with Ramachandaran plot having 100% residues of NadE in allowed region and overall quality factor of 81.75 using ERRAT tool. High throughput screening and Vina resulted in two potential hits against NadE such as ZINC83103551 and ZINC38008121. These molecules showed lowest binding energy of -5.7kcalmol-1 and high stability in the binding pockets during molecular dynamics simulation analysis. The similar approach for target identification could be applied for clinical strains of other pathogenic microbes.

Homology modeling and protein engineering of alkane monooxygenase in Burkholderia thailandensis MSMB121: in silico insights

AbstractThe degradation of hydrocarbons plays an important role in the eco-balancing of petroleum products, pesticides and other toxic products in the environment. The degradation of hydrocarbons by microbes such as Geobacillus thermodenitrificans, Burkhulderia, Gordonia sp. and Acinetobacter sp. has been studied intensively in the literature. The present study focused on the in silico protein engineering of alkane monooxygenase (ladA)—a protein involved in the alkane degradation pathway. We demonstrated the improvement in substrate binding energy with engineered ladA in Burkholderia thailandensis MSMB121. We identified an ortholog of ladA monooxygenase found in B. thailandensis MSMB121, and showed it to be an enzyme involved in an alkane degradation pathway studied extensively in Geobacillus thermodenitrificans. Homology modeling of the three-dimensional structure of ladA was performed with a crystal structure (protein databank ID: 3B9N) as a template in MODELLER 9v11, and further validated using PROCHECK, VERIFY-3D and WHATIF tools. Specific amino acids were substituted in the region corresponding to amino acids 305–370 of ladA protein, resulting in an enhancement of binding energy in different alkane chain molecules as compared to wild protein structures in the docking experiments. The substrate binding energy with the protein was calculated using Vina (Implemented in VEGAZZ). Molecular dynamics simulations were performed to study the dynamics of different alkane chain molecules inside the binding pockets of wild and mutated ladA. Here, we hypothesize an improvement in binding energies and accessibility of substrates towards engineered ladA enzyme, which could be further facilitated for wet laboratory-based experiments for validation of the alkane degradation pathway in this organism. This abstract shows homology based 3-D structure of alkane monooxygenase in Burkholderia thailandensis MSMB121, it was further validated using Ramachandaran plot, Whatif and verify-3D. Both wild and mutated alkane monooxygenase were docked with long chain alkane molecules (c16-c36) using Vina. Finally, a plot for RMSD of molecular dynamics simulations was generated using GROMACS showing fluctutations in wild and mutated protein structures

Feature selection for siRNA efficacy prediction using natural computation

RNAi is a naturally occurring, highly conserved phenomenon of RNA mediated gene silencing among the multicellular organisms. Currently, RNAi has been successfully applied in functional genomics, therapeutics and new drug target identification in mammals and other eukaryotes. The uniqueness lies in sequence specific gene knock down which made RNAi an indispensible technology.

PSO - SVM Based Classifiers: A Comparative Approach

Evolutionary and natural computing techniques have been drawn considerable interest for analyzing large datasets with large number of features. Various flavors of Particle Swarm Optimization (PSO) have been applied in the various research applications like Control and Automation, Function Optimization, Dimensionality Reduction, classification. In the present work, we have applied the SVM based classifier along with Novel PSO and Binary PSO on Huesken dataset of siRNA features as well as on nine other benchmark dataset and achieved results are quite satisfactory. The results of our study have been compared with other results available in the literature.

Identification of Phosphoribosyl-AMP cyclohydrolase, as drug target and its inhibitors in Brucella melitensis bv. 1 16M using metabolic pathway analysis.

Brucella melitensis is a pathogenic Gram-negative bacterium which is known for causing zoonotic diseases (Brucellosis). The organism is highly contagious and has been reported to be used as bioterrorism agent against humans. Several antibiotics and vaccines have been developed but these antibiotics have exhibited the sign of antibiotic resistance or ineffective at lower concentrations, which imposes an urgent need to identify the novel drugs/drug targets against this organism. In this work, metabolic pathways analysis has been performed with different filters such as non-homology with humans, essentially of genes and choke point analysis, leading to identification of novel drug targets. A total of 18 potential drug target proteins were filtered out and used to develop the high confidence protein–protein interaction network The Phosphoribosyl-AMP cyclohydrolase (HisI) protein has been identified as potential drug target on the basis of topological parameters. Further, a homology model of (HisI) protein has been developed using Modeller with multiple template (1W6Q (48%), 1ZPS (55%), and 2ZKN (48%)) approach and validated using PROCHECK and Verify3D. The virtual high throughput screening (vHTS) using DockBlaster tool has been performed against 16,11,889 clean fragments from ZINC database. Top 500 molecules from DockBlaster were docked using Vina. The docking analysis resulted in ZINC04880153 showing the lowest binding energy (−9.1 kcal/mol) with the drug target. The molecular dynamics study of the complex HisI-ZINC04880153 was conducted to analyze the stability and fluctuation of ligand within the binding pocket of HisI. The identified ligand could be analyzed in the wet-lab based experiments for future drug discovery.

A recursive PSO scheme for gene selection in microarray data

Abstract In DNA microarray datasets, the number of genes are very large, typically in thousands while the number of samples are in hundreds. This raises the issue of generalization in the classification process. Gene selection plays a significant role in improving the accuracy. In this paper, we have proposed a recursive particle swarm optimization approach (PSO) for gene selection. The proposed method refines the feature (gene) space from a very coarse level to a fine-grained one at each recursive step of the algorithm without degrading the accuracy. In addition, we have integrated various filter based ranking methods with the proposed recursive PSO approach. We also propose to use linear support vector machine weight vector to serve as initial gene pool selection. We evaluate our method on five publicly available benchmark microarray datasets. Our approach selects only a small number of genes while yielding substantial improvements in accuracy over state-of-the-art evolutionary methods.

Max-Margin feature selection

Many machine learning applications such as in vision, biology and social networking deal with data in high dimensions. Feature selection is typically employed to select a subset of features which im- proves generalization accuracy as well as reduces the computational cost of learning the model. One of the criteria used for feature selection is to jointly minimize the redundancy and maximize the rele- vance of the selected features. In this paper, we formulate the task of feature selection as a one class SVM problem in a space where features correspond to the data points and instances correspond to the dimensions. The goal is to look for a representative subset of the features (support vectors) which describes the boundary for the region where the set of the features (data points) exists. This leads to a joint optimization of relevance and redundancy in a principled max-margin framework. Additionally, our formulation enables us to leverage existing techniques for optimizing the SVM objective resulting in highly computationally efficient solutions for the task of feature selection. Specifically, we employ the dual coordinate descent algorithm (Hsieh et al., 2008), originally proposed for SVMs, for our formulation. We use a sparse representation to deal with data in very high dimensions. Experiments on seven publicly available benchmark datasets from a variety of domains show that our approach results in orders of magnitude faster solutions even while retaining the same level of accuracy compared to the state of the art feature selection techniques.

Variance Based Particle Swarm Optimization for Function Optimization and Feature Selection

Soft computing based techniques have been widely used in multi-objective optimization problems such as multi-modal function optimization, control and automation, network routing and feature selection etc. Feature Selection (FS) in high dimensional data can be modeled as multi-objective optimization problem to reduce the number of features while improving the overall accuracy. Generally, the traditional local optimization methods may not achieve this twin goal as there are many locally optimal solutions. Recently, various flavors of Particle Swarm Optimization (PSO) have been successfully applied for function optimization. The main issue in these variants of PSO is that it gets stuck in local optimum.

Homology modelling and molecular dynamics simulations of a protein serine/threonine phosphatase stp1 in Staphylococcus aureus N315: a potential drug target

The prevalence of methicillin-resistant Staphylococcus aureus and vanomycin intermediate S. aureus infections is on the rise, globally. This poses a huge challenge due to limited therapeutic options and the limited number of bacterial-specific drug targets available for due conservation with the human host. A serine/threonine phosphatase/kinase stp1/stk1 phospho-signalling system in S. aureus, which is just beginning to be understood, has been shown to be of importance in virulence and susceptibility to glycopeptide antibiotics. In this study, 3D structure of stp1 (clinical strain of S. aureus N315) was predicted using a homology modelling tool MODELLER. The validation of the predicted model was done using various tools such as PROCHECK, ERRAT, VERIFY-3D and ProSA. Molecular dynamics (MD) study was carried out using GROMACS to refine the least energy model generated from MODELLER9v11 and it was compared with the template. The template used was the crystal structure of serine/threonine phosphatase stp1 in Streptoccocus agalactiae (Protein Data Bank ID: 2PK0) with 38% identity with the query. Various validation tools showed the quality of the model generated using MODELLER. PROCHECK predicted 100% residues in the allowed region, ERRAT with overall quality factor of 76.47, VERIFY-3D with average score of >0.2 in 81.78% of residues, WHATIF with packaging quality score of > − 5 for all residues and ProSA with Z-score of − 7.02. MD simulation of the protein showed some fluctuations in the aqueous environment and changes in the ligand binding residues after simulation. The availability of the 3D-structural information of a viable drug target in S. aureus stp1 is expected to facilitate structure–activity relationship and interactions with proteins.