Mehdi Pirooznia

A comparative study of different machine learning methods on microarray gene expression data

BackgroundSeveral classification and feature selection methods have been studied for the identification of differentially expressed genes in microarray data. Classification methods such as SVM, RBF Neural Nets, MLP Neural Nets, Bayesian, Decision Tree and Random Forrest methods have been used in recent studies. The accuracy of these methods has been calculated with validation methods such as v-fold validation. However there is lack of comparison between these methods to find a better framework for classification, clustering and analysis of microarray gene expression results.ResultsIn this study, we compared the efficiency of the classification methods including; SVM, RBF Neural Nets, MLP Neural Nets, Bayesian, Decision Tree and Random Forrest methods. The v-fold cross validation was used to calculate the accuracy of the classifiers. Some of the common clustering methods including K-means, DBC, and EM clustering were applied to the datasets and the efficiency of these methods have been analysed. Further the efficiency of the feature selection methods including support vector machine recursive feature elimination (SVM-RFE), Chi Squared, and CSF were compared. In each case these methods were applied to eight different binary (two class) microarray datasets. We evaluated the class prediction efficiency of each gene list in training and test cross-validation using supervised classifiers.ConclusionsWe presented a study in which we compared some of the common used classification, clustering, and feature selection methods. We applied these methods to eight publicly available datasets, and compared how these methods performed in class prediction of test datasets. We reported that the choice of feature selection methods, the number of genes in the gene list, the number of cases (samples) substantially influence classification success. Based on features chosen by these methods, error rates and accuracy of several classification algorithms were obtained. Results revealed the importance of feature selection in accurately classifying new samples and how an integrated feature selection and classification algorithm is performing and is capable of identifying significant genes.

GeneVenn - A web application for comparing gene lists using Venn diagrams.

Numerous methods are available to compare results of multiple microarray studies. One of the simplest but most effective of these procedures is to examine the overlap of resulting gene lists in a Venn diagram. Venn diagrams are graphical ways of representing interactions among sets to display information that can be read easily. Here we propose a simple but effective web application creating Venn diagrams from two or three gene lists. Each gene in the group list has link to the related information in NCBIs Entrez Nucleotide database. Availability GeneVenn is available for free at http://mcbc.usm.edu/genevenn/

Cloning, Analysis and Functional Annotation of Expressed Sequence tags from the Earthworm Eisenia fetida

BackgroundEisenia fetida, commonly known as red wiggler or compost worm, belongs to the Lumbricidae family of the Annelida phylum. Little is known about its genome sequence although it has been extensively used as a test organism in terrestrial ecotoxicology. In order to understand its gene expression response to environmental contaminants, we cloned 4032 cDNAs or expressed sequence tags (ESTs) from two E. fetida libraries enriched with genes responsive to ten ordnance related compounds using suppressive subtractive hybridization-PCR.ResultsA total of 3144 good quality ESTs (GenBank dbEST accession number EH669363–EH672369 and EL515444–EL515580) were obtained from the raw clone sequences after cleaning. Clustering analysis yielded 2231 unique sequences including 448 contigs (from 1361 ESTs) and 1783 singletons. Comparative genomic analysis showed that 743 or 33% of the unique sequences shared high similarity with existing genes in the GenBank nr database. Provisional function annotation assigned 830 Gene Ontology terms to 517 unique sequences based on their homology with the annotated genomes of four model organisms Drosophila melanogaster, Mus musculus, Saccharomyces cerevisiae, and Caenorhabditis elegans. Seven percent of the unique sequences were further mapped to 99 Kyoto Encyclopedia of Genes and Genomes pathways based on their matching Enzyme Commission numbers. All the information is stored and retrievable at a highly performed, web-based and user-friendly relational database called EST model database or ESTMD version 2.ConclusionThe ESTMD containing the sequence and annotation information of 4032 E. fetida ESTs is publicly accessible at http://mcbc.usm.edu/estmd/.

Transcriptomic analysis of RDX and TNT interactive sublethal effects in the earthworm Eisenia fetida

BackgroundExplosive compounds such as TNT and RDX are recalcitrant contaminants often found co-existing in the environment. In order to understand the joint effects of TNT and RDX on earthworms, an important ecological and bioindicator species at the molecular level, we sampled worms (Eisenia fetida) exposed singly or jointly to TNT (50 mg/kg soil) and RDX (30 mg/kg soil) for 28 days and profiled gene expression in an interwoven loop designed microarray experiment using a 4k-cDNA array. Lethality, growth and reproductive endpoints were measured.ResultsSublethal doses of TNT and RDX had no significant effects on the survival and growth of earthworms, but significantly reduced cocoon and juvenile counts. The mixture exhibited more pronounced reproductive toxicity than each single compound, suggesting an additive interaction between the two compounds. In comparison with the controls, we identified 321 differentially expressed transcripts in TNT treated worms, 32 in RDX treated worms, and only 6 in mixture treated worms. Of the 329 unique differentially expressed transcripts, 294 were affected only by TNT, 24 were common to both TNT and RDX treatments, and 3 were common to all treatments. The reduced effects on gene expression in the mixture exposure suggest that RDX might interact in an antagonistic manner with TNT at the gene expression level. The disagreement between gene expression and reproduction results may be attributed to sampling time, absence of known reproduction-related genes, and lack of functional information for many differentially expressed transcripts. A gene potentially related to reproduction (echinonectin) was significantly depressed in TNT or RDX exposed worms and may be linked to reduced fecundity.ConclusionsSublethal doses of TNT and RDX affected many biological pathways from innate immune response to oogenesis, leading to reduced reproduction without affecting survival and growth. A complex interaction between mixtures of RDX and TNT was observed at the gene expression level that requires further study of the dynamics of gene expression and reproductive activities in E. fetida. These efforts will be essential to gain an understanding of the additive reproductive toxicity between RDX and TNT.

SVM Classifier – a comprehensive java interface for support vector machine classification of microarray data

MotivationGraphical user interface (GUI) software promotes novelty by allowing users to extend the functionality. SVM Classifier is a cross-platform graphical application that handles very large datasets well. The purpose of this study is to create a GUI application that allows SVM users to perform SVM training, classification and prediction.ResultsThe GUI provides user-friendly access to state-of-the-art SVM methods embodied in the LIBSVM implementation of Support Vector Machine. We implemented the java interface using standard swing libraries.We used a sample data from a breast cancer study for testing classification accuracy. We achieved 100% accuracy in classification among the BRCA1–BRCA2 samples with RBF kernel of SVM.ConclusionWe have developed a java GUI application that allows SVM users to perform SVM training, classification and prediction. We have demonstrated that support vector machines can accurately classify genes into functional categories based upon expression data from DNA microarray hybridization experiments. Among the different kernel functions that we examined, the SVM that uses a radial basis kernel function provides the best performance.The SVM Classifier is available at http://mfgn.usm.edu/ebl/svm/.

RiboaptDB: A Comprehensive Database of Ribozymes and Aptamers

BackgroundCatalytic RNA molecules are called ribozymes. The aptamers are DNA or RNA molecules that have been selected from vast populations of random sequences, through a combinatorial approach known as SELEX. The selected oligo-nucleotide sequences (~200 bp in length) have the ability to recognize a broad range of specific ligands by forming binding pockets. These novel aptamer sequences can bind to nucleic acids, proteins or small organic and inorganic chemical compounds and have many potential uses in medicine and technology.ResultsThe comprehensive sequence information on aptamers and ribozymes that have been generated by in vitro selection methods are included in this RiboaptDB database. Such types of unnatural data generated by in vitro methods are not available in the public natural sequence databases such as GenBank and EMBL. The amount of sequence data generated by in vitro selection experiments has been accumulating exponentially. There are 370 artificial ribozyme sequences and 3842 aptamer sequences in the total 4212 sequences from 423 citations in this RiboaptDB. We included general search feature, and individual feature wise search, user submission form for new data through online and also local BLAST search.ConclusionThis database, besides serving as a storehouse of sequences that may have diagnostic or therapeutic utility in medicine, provides valuable information for computational and theoretical biologists. The RiboaptDB is extremely useful for garnering information about in vitro selection experiments as a whole and for better understanding the distribution of functional nucleic acids in sequence space. The database is updated regularly and is publicly available at http://mfgn.usm.edu/ebl/riboapt/.

Batch Blast Extractor: an automated blastx parser application

MotivationBLAST programs are very efficient in finding similarities for sequences. However for large datasets such as ESTs, manual extraction of the information from the batch BLAST output is needed. This can be time consuming, insufficient, and inaccurate. Therefore implementation of a parser application would be extremely useful in extracting information from BLAST outputs.ResultsWe have developed a java application, Batch Blast Extractor, with a user friendly graphical interface to extract information from BLAST output. The application generates a tab delimited text file that can be easily imported into any statistical package such as Excel or SPSS for further analysis. For each BLAST hit, the program obtains and saves the essential features from the BLAST output file that would allow further analysis. The program was written in Java and therefore is OS independent. It works on both Windows and Linux OS with java 1.4 and higher. It is freely available from: http://mcbc.usm.edu/BatchBlastExtractor/

ILOOP – a web application for two-channel microarray interwoven loop design

Microarray technology is widely applied to address complex scientific questions. However, there remain fundamental issues on how to design experiments to ensure that the resulting data enables robust statistical analysis. Interwoven loop design has several advantages over other designs. However it suffers in the complexity of design. We have implemented an online web application which allows users to find optimal loop designs for two-color microarray experiments. Given a number of conditions (such as treatments or time points) and replicates, the application will find the best possible design of the experiment and output experimental parameters. It is freely available from http://mcbc.usm.edu/iloop.

Gene expression analysis of CL-20-induced reversible neurotoxicity reveals GABA(A) receptors as potential targets in the earthworm Eisenia fetida.

The earthworm Eisenia fetida is one of the most used species in standardized soil ecotoxicity tests. End points such as survival, growth, and reproduction are eco-toxicologically relevant but provide little mechanistic insight into toxicity pathways, especially at the molecular level. Here we apply a toxicogenomic approach to investigate the mode of action underlying the reversible neurotoxicity of hexanitrohexaazaisowurtzitane (CL-20), a cyclic nitroamine explosives compound. We developed an E. fetida-specific shotgun microarray targeting 15119 unique E. fetida transcripts. Using this array we profiled gene expression in E. fetida in response to exposure to CL-20. Eighteen earthworms were exposed for 6 days to 0.2 μg/cm(2) of CL-20 on filter paper, half of which were allowed to recover in a clean environment for 7 days. Nine vehicle control earthworms were sacrificed at days 6 and 13, separately. Electrophysiological measurements indicated that the conduction velocity of earthworm medial giant nerve fiber decreased significantly after 6-day exposure to CL-20, but was restored after 7 days of recovery. Total RNA was isolated from the four treatment groups including 6-day control, 6-day exposed, 13-day control, and 13-day exposed (i.e., 6-day exposure followed by 7-day recovery), and was hybridized to the 15K shotgun oligo array. Statistical and bioinformatic analyses suggest that CL-20 initiated neurotoxicity by noncompetitively blocking the ligand-gated GABA(A) receptor ion channel, leading to altered expression of genes involved in GABAergic, cholinergic, and Agrin-MuSK pathways. In the recovery phase, expression of affected genes returned to normality, possibly as a result of autophagy and CL-20 dissociation/metabolism. This study provides significant insights into potential mechanisms of CL-20-induced neurotoxicity and the recovery of earthworms from transient neurotoxicity stress.

Efficiency of Hybrid Normalization of Microarray Gene Expression: A Simulation Study

Microarray experiments data contain errors, from various sources. The process of identifying and adjusting of systematic variation in intensities between samples on the same slide is known as normalization. We implemented a Java application with graphical user interface that allows easy evaluation of errors and the role of hybrid normalization methods to remove the systematic errors from the experiments data. We calculated true hybridization intensity then applied different additive and multiplicative errors and computed the normalized intensity logged ratio using background subtraction, self-normalization method and dye-flip technique. Results show the efficiency of hybrid normalization method in order to remove error rates from the intensity values.